diff --git a/.dockerignore b/.dockerignore
index 97e8aa34..2c394c48 100644
--- a/.dockerignore
+++ b/.dockerignore
@@ -1,4 +1,6 @@
 .idea
+.github
+.vscode
 models
 examples/chatbot-ui/models
 examples/rwkv/models
diff --git a/.github/workflows/test-extra.yml b/.github/workflows/test-extra.yml
index 7689f06d..7705783e 100644
--- a/.github/workflows/test-extra.yml
+++ b/.github/workflows/test-extra.yml
@@ -32,8 +32,9 @@ jobs:
              sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
              sudo apt-get update && \
              sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
           sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools
           
           sudo rm -rfv /usr/bin/conda || true
 
@@ -61,8 +62,9 @@ jobs:
              sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
              sudo apt-get update && \
              sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
           sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools
           
           sudo rm -rfv /usr/bin/conda || true
 
@@ -90,8 +92,9 @@ jobs:
              sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
              sudo apt-get update && \
              sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
           sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools
           
           sudo rm -rfv /usr/bin/conda || true
 
@@ -120,8 +123,9 @@ jobs:
              sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
              sudo apt-get update && \
              sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
           sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools
           
           sudo rm -rfv /usr/bin/conda || true
 
@@ -151,8 +155,9 @@ jobs:
   #            sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
   #            sudo apt-get update && \
   #            sudo apt-get install -y conda
-  #         sudo apt-get install -y ca-certificates cmake curl patch
+  #         sudo apt-get install -y ca-certificates cmake curl patch python3-pip
   #         sudo apt-get install -y libopencv-dev
+  #         pip install --user grpcio-tools
           
   #         sudo rm -rfv /usr/bin/conda || true
 
@@ -222,8 +227,9 @@ jobs:
   #            sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
   #            sudo apt-get update && \
   #            sudo apt-get install -y conda
-  #         sudo apt-get install -y ca-certificates cmake curl patch
+  #         sudo apt-get install -y ca-certificates cmake curl patch python3-pip
   #         sudo apt-get install -y libopencv-dev
+  #         pip install --user grpcio-tools
           
   #         sudo rm -rfv /usr/bin/conda || true
 
@@ -254,8 +260,9 @@ jobs:
   #            sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
   #            sudo apt-get update && \
   #            sudo apt-get install -y conda
-  #         sudo apt-get install -y ca-certificates cmake curl patch
+  #         sudo apt-get install -y ca-certificates cmake curl patch python3-pip
   #         sudo apt-get install -y libopencv-dev
+  #         pip install --user grpcio-tools
   #         sudo rm -rfv /usr/bin/conda || true
   #     - name: Test vllm
   #       run: |
@@ -280,8 +287,9 @@ jobs:
              sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
              sudo apt-get update && \
              sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
-          sudo apt-get install -y libopencv-dev    
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools
           sudo rm -rfv /usr/bin/conda || true
       - name: Test vall-e-x
         run: |
@@ -307,7 +315,8 @@ jobs:
              sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
              sudo apt-get update && \
              sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch espeak espeak-ng          
+          sudo apt-get install -y ca-certificates cmake curl patch espeak espeak-ng python3-pip
+          pip install --user grpcio-tools
           sudo rm -rfv /usr/bin/conda || true
 
       - name: Test coqui
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 02093b3f..46c4e065 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -70,17 +70,27 @@ jobs:
       - name: Dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install build-essential ffmpeg
+          sudo apt-get install build-essential curl ffmpeg
           curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
              sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
+             gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
              sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
              sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
              sudo apt-get update && \
              sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
+          sudo apt-get install -y ca-certificates cmake patch python3-pip unzip
           sudo apt-get install -y libopencv-dev
-          
+
+          curl -L -s https://github.com/protocolbuffers/protobuf/releases/download/v26.1/protoc-26.1-linux-x86_64.zip -o protoc.zip && \
+          unzip -j -d /usr/local/bin protoc.zip bin/protoc && \
+          rm protoc.zip
+
+          go install google.golang.org/protobuf/cmd/protoc-gen-go@latest
+          go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@latest
+
+          # The python3-grpc-tools package in 22.04 is too old
+          pip install --user grpcio-tools
+
           sudo rm -rfv /usr/bin/conda || true
           PATH=$PATH:/opt/conda/bin make -C backend/python/sentencetransformers
 
@@ -89,7 +99,7 @@ jobs:
           GO_TAGS="tts" make -C sources/go-piper piper.o && \
           sudo cp -rfv sources/go-piper/piper-phonemize/pi/lib/. /usr/lib/ && \
           # Pre-build stable diffusion before we install a newer version of abseil (not compatible with stablediffusion-ncn)
-          GO_TAGS="stablediffusion tts" GRPC_BACKENDS=backend-assets/grpc/stablediffusion make build
+          PATH="$PATH:/root/go/bin" GO_TAGS="stablediffusion tts" GRPC_BACKENDS=backend-assets/grpc/stablediffusion make build
       - name: Cache grpc
         id: cache-grpc
         uses: actions/cache@v4
@@ -108,7 +118,7 @@ jobs:
           cd grpc && cd cmake/build && sudo make --jobs 5 install
       - name: Test
         run: |
-          GO_TAGS="stablediffusion tts" make --jobs 5 --output-sync=target test
+          PATH="$PATH:/root/go/bin" GO_TAGS="stablediffusion tts" make --jobs 5 --output-sync=target test
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
         uses: mxschmitt/action-tmate@v3
@@ -186,7 +196,8 @@ jobs:
         run: go version
       - name: Dependencies
         run: |
-          brew install protobuf grpc make
+          brew install protobuf grpc make protoc-gen-go protoc-gen-go-grpc
+          pip install --user grpcio-tools
       - name: Test
         run: |
           export C_INCLUDE_PATH=/usr/local/include
diff --git a/.gitignore b/.gitignore
index b48f7391..f1f860e9 100644
--- a/.gitignore
+++ b/.gitignore
@@ -39,3 +39,8 @@ backend-assets/*
 !backend-assets/.keep
 prepare
 /ggml-metal.metal
+
+# Protobuf generated files
+*.pb.go
+*pb2.py
+*pb2_grpc.py
diff --git a/Dockerfile b/Dockerfile
index 5fb6230c..d0217d50 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -20,12 +20,25 @@ ENV EXTERNAL_GRPC_BACKENDS="coqui:/build/backend/python/coqui/run.sh,huggingface
 ARG GO_TAGS="stablediffusion tinydream tts"
 
 RUN apt-get update && \
-    apt-get install -y ca-certificates curl patch pip cmake git && apt-get clean
+    apt-get install -y ca-certificates curl python3-pip unzip && apt-get clean
 
 # Install Go
 RUN curl -L -s https://go.dev/dl/go$GO_VERSION.linux-$TARGETARCH.tar.gz | tar -C /usr/local -xz
 ENV PATH $PATH:/usr/local/go/bin
 
+# Install grpc compilers
+ENV PATH $PATH:/root/go/bin
+RUN go install google.golang.org/protobuf/cmd/protoc-gen-go@latest && \
+    go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@latest
+
+# Install protobuf (the version in 22.04 is too old)
+RUN curl -L -s https://github.com/protocolbuffers/protobuf/releases/download/v26.1/protoc-26.1-linux-x86_64.zip -o protoc.zip && \
+    unzip -j -d /usr/local/bin protoc.zip bin/protoc && \
+    rm protoc.zip
+
+# Install grpcio-tools (the version in 22.04 is too old)
+RUN pip install --user grpcio-tools
+
 COPY --chmod=644 custom-ca-certs/* /usr/local/share/ca-certificates/
 RUN update-ca-certificates
 
@@ -68,7 +81,8 @@ RUN test -n "$TARGETARCH" \
 
 FROM requirements-core as requirements-extras
 
-RUN curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
+RUN apt install -y gpg && \
+    curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
     install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
     gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
     echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list && \
@@ -100,7 +114,7 @@ ENV MAKEFLAGS=${MAKEFLAGS}
 WORKDIR /build
 
 RUN apt-get update && \
-    apt-get install -y g++ cmake git && \
+    apt-get install -y build-essential cmake git  && \
     apt-get clean && \
     rm -rf /var/lib/apt/lists/*
 
@@ -133,6 +147,12 @@ WORKDIR /build
 COPY . .
 COPY .git .
 RUN echo "GO_TAGS: $GO_TAGS"
+
+RUN apt-get update && \
+    apt-get install -y build-essential cmake git  && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
 RUN make prepare
 
 # If we are building with clblas support, we need the libraries for the builds
@@ -191,6 +211,11 @@ RUN if [ "${BUILD_TYPE}" = "clblas" ]; then \
     apt-get clean \
     ; fi
 
+RUN apt-get update && \
+    apt-get install -y cmake git  && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
 WORKDIR /build
 
 # we start fresh & re-copy all assets because `make build` does not clean up nicely after itself
@@ -202,7 +227,7 @@ COPY . .
 COPY --from=builder /build/sources ./sources/
 COPY --from=grpc /build/grpc ./grpc/
 
-RUN make prepare-sources && cd /build/grpc/cmake/build && make install && rm -rf grpc
+RUN make prepare-sources && cd /build/grpc/cmake/build && make install && rm -rf /build/grpc
 
 # Copy the binary
 COPY --from=builder /build/local-ai ./
diff --git a/Makefile b/Makefile
index 0f6d8fd2..5932dfb2 100644
--- a/Makefile
+++ b/Makefile
@@ -289,10 +289,12 @@ clean: ## Remove build related file
 	rm -rf ./sources
 	rm -rf $(BINARY_NAME)
 	rm -rf release/
-	rm -rf backend-assets
+	rm -rf backend-assets/*
 	$(MAKE) -C backend/cpp/grpc clean
 	$(MAKE) -C backend/cpp/llama clean
 	$(MAKE) dropreplace
+	$(MAKE) protogen-clean
+	rmdir pkg/grpc/proto || true
 
 clean-tests:
 	rm -rf test-models
@@ -416,30 +418,136 @@ help: ## Show this help.
 		else if (/^## .*$$/) {printf "  ${CYAN}%s${RESET}\n", substr($$1,4)} \
 		}' $(MAKEFILE_LIST)
 
+.PHONY: protogen
 protogen: protogen-go protogen-python
 
+.PHONY: protogen-clean
+protogen-clean: protogen-go-clean protogen-python-clean
+
+.PHONY: protogen-go
 protogen-go:
+	mkdir -p pkg/grpc/proto
 	protoc -Ibackend/ --go_out=pkg/grpc/proto/ --go_opt=paths=source_relative --go-grpc_out=pkg/grpc/proto/ --go-grpc_opt=paths=source_relative \
     backend/backend.proto
 
-protogen-python:
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/sentencetransformers/ --grpc_python_out=backend/python/sentencetransformers/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/transformers/ --grpc_python_out=backend/python/transformers/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/transformers-musicgen/ --grpc_python_out=backend/python/transformers-musicgen/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/autogptq/ --grpc_python_out=backend/python/autogptq/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/exllama/ --grpc_python_out=backend/python/exllama/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/bark/ --grpc_python_out=backend/python/bark/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/diffusers/ --grpc_python_out=backend/python/diffusers/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/coqui/ --grpc_python_out=backend/python/coqui/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/vall-e-x/ --grpc_python_out=backend/python/vall-e-x/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/vllm/ --grpc_python_out=backend/python/vllm/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/petals/ --grpc_python_out=backend/python/petals/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/mamba/ --grpc_python_out=backend/python/mamba/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/exllama2/ --grpc_python_out=backend/python/exllama2/ backend/backend.proto
+.PHONY: protogen-go-clean
+protogen-go-clean:
+	$(RM) pkg/grpc/proto/backend.pb.go pkg/grpc/proto/backend_grpc.pb.go
+	$(RM) bin/*
+
+.PHONY: protogen-python
+protogen-python: autogptq-protogen bark-protogen coqui-protogen diffusers-protogen exllama-protogen exllama2-protogen mamba-protogen petals-protogen sentencetransformers-protogen transformers-protogen transformers-musicgen-protogen vall-e-x-protogen vllm-protogen
+
+.PHONY: protogen-python-clean
+protogen-python-clean: autogptq-protogen-clean bark-protogen-clean coqui-protogen-clean diffusers-protogen-clean exllama-protogen-clean exllama2-protogen-clean mamba-protogen-clean petals-protogen-clean sentencetransformers-protogen-clean transformers-protogen-clean transformers-musicgen-protogen-clean vall-e-x-protogen-clean vllm-protogen-clean
+
+.PHONY: autogptq-protogen
+autogptq-protogen:
+	$(MAKE) -C backend/python/autogptq protogen
+
+.PHONY: autogptq-protogen-clean
+autogptq-protogen-clean:
+	$(MAKE) -C backend/python/autogptq protogen-clean
+
+.PHONY: bark-protogen
+bark-protogen:
+	$(MAKE) -C backend/python/bark protogen
+
+.PHONY: bark-protogen-clean
+bark-protogen-clean:
+	$(MAKE) -C backend/python/bark protogen-clean
+
+.PHONY: coqui-protogen
+coqui-protogen:
+	$(MAKE) -C backend/python/coqui protogen
+
+.PHONY: coqui-protogen-clean
+coqui-protogen-clean:
+	$(MAKE) -C backend/python/coqui protogen-clean
+
+.PHONY: diffusers-protogen
+diffusers-protogen:
+	$(MAKE) -C backend/python/diffusers protogen
+
+.PHONY: diffusers-protogen-clean
+diffusers-protogen-clean:
+	$(MAKE) -C backend/python/diffusers protogen-clean
+
+.PHONY: exllama-protogen
+exllama-protogen:
+	$(MAKE) -C backend/python/exllama protogen
+
+.PHONY: exllama-protogen-clean
+exllama-protogen-clean:
+	$(MAKE) -C backend/python/exllama protogen-clean
+
+.PHONY: exllama2-protogen
+exllama2-protogen:
+	$(MAKE) -C backend/python/exllama2 protogen
+
+.PHONY: exllama2-protogen-clean
+exllama2-protogen-clean:
+	$(MAKE) -C backend/python/exllama2 protogen-clean
+
+.PHONY: mamba-protogen
+mamba-protogen:
+	$(MAKE) -C backend/python/mamba protogen
+
+.PHONY: mamba-protogen-clean
+mamba-protogen-clean:
+	$(MAKE) -C backend/python/mamba protogen-clean
+
+.PHONY: petals-protogen
+petals-protogen:
+	$(MAKE) -C backend/python/petals protogen
+
+.PHONY: petals-protogen-clean
+petals-protogen-clean:
+	$(MAKE) -C backend/python/petals protogen-clean
+
+.PHONY: sentencetransformers-protogen
+sentencetransformers-protogen:
+	$(MAKE) -C backend/python/sentencetransformers protogen
+
+.PHONY: sentencetransformers-protogen-clean
+sentencetransformers-protogen-clean:
+	$(MAKE) -C backend/python/sentencetransformers protogen-clean
+
+.PHONY: transformers-protogen
+transformers-protogen:
+	$(MAKE) -C backend/python/transformers protogen
+
+.PHONY: transformers-protogen-clean
+transformers-protogen-clean:
+	$(MAKE) -C backend/python/transformers protogen-clean
+
+.PHONY: transformers-musicgen-protogen
+transformers-musicgen-protogen:
+	$(MAKE) -C backend/python/transformers-musicgen protogen
+
+.PHONY: transformers-musicgen-protogen-clean
+transformers-musicgen-protogen-clean:
+	$(MAKE) -C backend/python/transformers-musicgen protogen-clean
+
+.PHONY: vall-e-x-protogen
+vall-e-x-protogen:
+	$(MAKE) -C backend/python/vall-e-x protogen
+
+.PHONY: vall-e-x-protogen-clean
+vall-e-x-protogen-clean:
+	$(MAKE) -C backend/python/vall-e-x protogen-clean
+
+.PHONY: vllm-protogen
+vllm-protogen:
+	$(MAKE) -C backend/python/vllm protogen
+
+.PHONY: vllm-protogen-clean
+vllm-protogen-clean:
+	$(MAKE) -C backend/python/vllm protogen-clean
 
 ## GRPC
 # Note: it is duplicated in the Dockerfile
-prepare-extra-conda-environments:
+prepare-extra-conda-environments: protogen-python
 	$(MAKE) -C backend/python/autogptq
 	$(MAKE) -C backend/python/bark
 	$(MAKE) -C backend/python/coqui
@@ -454,7 +562,7 @@ prepare-extra-conda-environments:
 	$(MAKE) -C backend/python/petals
 	$(MAKE) -C backend/python/exllama2
 
-prepare-test-extra:
+prepare-test-extra: protogen-python
 	$(MAKE) -C backend/python/transformers
 	$(MAKE) -C backend/python/diffusers
 
@@ -478,7 +586,7 @@ backend-assets/gpt4all: sources/gpt4all sources/gpt4all/gpt4all-bindings/golang/
 	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dylib backend-assets/gpt4all/ || true
 	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dll backend-assets/gpt4all/ || true
 
-backend-assets/grpc: replace
+backend-assets/grpc: protogen-go replace
 	mkdir -p backend-assets/grpc
 
 backend-assets/grpc/bert-embeddings: sources/go-bert sources/go-bert/libgobert.a backend-assets/grpc
diff --git a/backend/backend_grpc.pb.go b/backend/backend_grpc.pb.go
deleted file mode 100644
index 5c97691d..00000000
--- a/backend/backend_grpc.pb.go
+++ /dev/null
@@ -1,457 +0,0 @@
-// Code generated by protoc-gen-go-grpc. DO NOT EDIT.
-// versions:
-// - protoc-gen-go-grpc v1.2.0
-// - protoc             v4.23.4
-// source: backend/backend.proto
-
-package proto
-
-import (
-	context "context"
-	grpc "google.golang.org/grpc"
-	codes "google.golang.org/grpc/codes"
-	status "google.golang.org/grpc/status"
-)
-
-// This is a compile-time assertion to ensure that this generated file
-// is compatible with the grpc package it is being compiled against.
-// Requires gRPC-Go v1.32.0 or later.
-const _ = grpc.SupportPackageIsVersion7
-
-// BackendClient is the client API for Backend service.
-//
-// For semantics around ctx use and closing/ending streaming RPCs, please refer to https://pkg.go.dev/google.golang.org/grpc/?tab=doc#ClientConn.NewStream.
-type BackendClient interface {
-	Health(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*Reply, error)
-	Predict(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*Reply, error)
-	LoadModel(ctx context.Context, in *ModelOptions, opts ...grpc.CallOption) (*Result, error)
-	PredictStream(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (Backend_PredictStreamClient, error)
-	Embedding(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*EmbeddingResult, error)
-	GenerateImage(ctx context.Context, in *GenerateImageRequest, opts ...grpc.CallOption) (*Result, error)
-	AudioTranscription(ctx context.Context, in *TranscriptRequest, opts ...grpc.CallOption) (*TranscriptResult, error)
-	TTS(ctx context.Context, in *TTSRequest, opts ...grpc.CallOption) (*Result, error)
-	TokenizeString(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*TokenizationResponse, error)
-	Status(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*StatusResponse, error)
-}
-
-type backendClient struct {
-	cc grpc.ClientConnInterface
-}
-
-func NewBackendClient(cc grpc.ClientConnInterface) BackendClient {
-	return &backendClient{cc}
-}
-
-func (c *backendClient) Health(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*Reply, error) {
-	out := new(Reply)
-	err := c.cc.Invoke(ctx, "/backend.Backend/Health", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) Predict(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*Reply, error) {
-	out := new(Reply)
-	err := c.cc.Invoke(ctx, "/backend.Backend/Predict", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) LoadModel(ctx context.Context, in *ModelOptions, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, "/backend.Backend/LoadModel", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) PredictStream(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (Backend_PredictStreamClient, error) {
-	stream, err := c.cc.NewStream(ctx, &Backend_ServiceDesc.Streams[0], "/backend.Backend/PredictStream", opts...)
-	if err != nil {
-		return nil, err
-	}
-	x := &backendPredictStreamClient{stream}
-	if err := x.ClientStream.SendMsg(in); err != nil {
-		return nil, err
-	}
-	if err := x.ClientStream.CloseSend(); err != nil {
-		return nil, err
-	}
-	return x, nil
-}
-
-type Backend_PredictStreamClient interface {
-	Recv() (*Reply, error)
-	grpc.ClientStream
-}
-
-type backendPredictStreamClient struct {
-	grpc.ClientStream
-}
-
-func (x *backendPredictStreamClient) Recv() (*Reply, error) {
-	m := new(Reply)
-	if err := x.ClientStream.RecvMsg(m); err != nil {
-		return nil, err
-	}
-	return m, nil
-}
-
-func (c *backendClient) Embedding(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*EmbeddingResult, error) {
-	out := new(EmbeddingResult)
-	err := c.cc.Invoke(ctx, "/backend.Backend/Embedding", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) GenerateImage(ctx context.Context, in *GenerateImageRequest, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, "/backend.Backend/GenerateImage", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) AudioTranscription(ctx context.Context, in *TranscriptRequest, opts ...grpc.CallOption) (*TranscriptResult, error) {
-	out := new(TranscriptResult)
-	err := c.cc.Invoke(ctx, "/backend.Backend/AudioTranscription", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) TTS(ctx context.Context, in *TTSRequest, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, "/backend.Backend/TTS", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) TokenizeString(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*TokenizationResponse, error) {
-	out := new(TokenizationResponse)
-	err := c.cc.Invoke(ctx, "/backend.Backend/TokenizeString", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) Status(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*StatusResponse, error) {
-	out := new(StatusResponse)
-	err := c.cc.Invoke(ctx, "/backend.Backend/Status", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-// BackendServer is the server API for Backend service.
-// All implementations must embed UnimplementedBackendServer
-// for forward compatibility
-type BackendServer interface {
-	Health(context.Context, *HealthMessage) (*Reply, error)
-	Predict(context.Context, *PredictOptions) (*Reply, error)
-	LoadModel(context.Context, *ModelOptions) (*Result, error)
-	PredictStream(*PredictOptions, Backend_PredictStreamServer) error
-	Embedding(context.Context, *PredictOptions) (*EmbeddingResult, error)
-	GenerateImage(context.Context, *GenerateImageRequest) (*Result, error)
-	AudioTranscription(context.Context, *TranscriptRequest) (*TranscriptResult, error)
-	TTS(context.Context, *TTSRequest) (*Result, error)
-	TokenizeString(context.Context, *PredictOptions) (*TokenizationResponse, error)
-	Status(context.Context, *HealthMessage) (*StatusResponse, error)
-	mustEmbedUnimplementedBackendServer()
-}
-
-// UnimplementedBackendServer must be embedded to have forward compatible implementations.
-type UnimplementedBackendServer struct {
-}
-
-func (UnimplementedBackendServer) Health(context.Context, *HealthMessage) (*Reply, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Health not implemented")
-}
-func (UnimplementedBackendServer) Predict(context.Context, *PredictOptions) (*Reply, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Predict not implemented")
-}
-func (UnimplementedBackendServer) LoadModel(context.Context, *ModelOptions) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method LoadModel not implemented")
-}
-func (UnimplementedBackendServer) PredictStream(*PredictOptions, Backend_PredictStreamServer) error {
-	return status.Errorf(codes.Unimplemented, "method PredictStream not implemented")
-}
-func (UnimplementedBackendServer) Embedding(context.Context, *PredictOptions) (*EmbeddingResult, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Embedding not implemented")
-}
-func (UnimplementedBackendServer) GenerateImage(context.Context, *GenerateImageRequest) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method GenerateImage not implemented")
-}
-func (UnimplementedBackendServer) AudioTranscription(context.Context, *TranscriptRequest) (*TranscriptResult, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method AudioTranscription not implemented")
-}
-func (UnimplementedBackendServer) TTS(context.Context, *TTSRequest) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method TTS not implemented")
-}
-func (UnimplementedBackendServer) TokenizeString(context.Context, *PredictOptions) (*TokenizationResponse, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method TokenizeString not implemented")
-}
-func (UnimplementedBackendServer) Status(context.Context, *HealthMessage) (*StatusResponse, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Status not implemented")
-}
-func (UnimplementedBackendServer) mustEmbedUnimplementedBackendServer() {}
-
-// UnsafeBackendServer may be embedded to opt out of forward compatibility for this service.
-// Use of this interface is not recommended, as added methods to BackendServer will
-// result in compilation errors.
-type UnsafeBackendServer interface {
-	mustEmbedUnimplementedBackendServer()
-}
-
-func RegisterBackendServer(s grpc.ServiceRegistrar, srv BackendServer) {
-	s.RegisterService(&Backend_ServiceDesc, srv)
-}
-
-func _Backend_Health_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(HealthMessage)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Health(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/Health",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Health(ctx, req.(*HealthMessage))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_Predict_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(PredictOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Predict(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/Predict",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Predict(ctx, req.(*PredictOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_LoadModel_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(ModelOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).LoadModel(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/LoadModel",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).LoadModel(ctx, req.(*ModelOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_PredictStream_Handler(srv interface{}, stream grpc.ServerStream) error {
-	m := new(PredictOptions)
-	if err := stream.RecvMsg(m); err != nil {
-		return err
-	}
-	return srv.(BackendServer).PredictStream(m, &backendPredictStreamServer{stream})
-}
-
-type Backend_PredictStreamServer interface {
-	Send(*Reply) error
-	grpc.ServerStream
-}
-
-type backendPredictStreamServer struct {
-	grpc.ServerStream
-}
-
-func (x *backendPredictStreamServer) Send(m *Reply) error {
-	return x.ServerStream.SendMsg(m)
-}
-
-func _Backend_Embedding_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(PredictOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Embedding(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/Embedding",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Embedding(ctx, req.(*PredictOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_GenerateImage_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(GenerateImageRequest)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).GenerateImage(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/GenerateImage",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).GenerateImage(ctx, req.(*GenerateImageRequest))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_AudioTranscription_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(TranscriptRequest)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).AudioTranscription(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/AudioTranscription",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).AudioTranscription(ctx, req.(*TranscriptRequest))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_TTS_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(TTSRequest)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).TTS(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/TTS",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).TTS(ctx, req.(*TTSRequest))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_TokenizeString_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(PredictOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).TokenizeString(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/TokenizeString",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).TokenizeString(ctx, req.(*PredictOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_Status_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(HealthMessage)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Status(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/Status",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Status(ctx, req.(*HealthMessage))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-// Backend_ServiceDesc is the grpc.ServiceDesc for Backend service.
-// It's only intended for direct use with grpc.RegisterService,
-// and not to be introspected or modified (even as a copy)
-var Backend_ServiceDesc = grpc.ServiceDesc{
-	ServiceName: "backend.Backend",
-	HandlerType: (*BackendServer)(nil),
-	Methods: []grpc.MethodDesc{
-		{
-			MethodName: "Health",
-			Handler:    _Backend_Health_Handler,
-		},
-		{
-			MethodName: "Predict",
-			Handler:    _Backend_Predict_Handler,
-		},
-		{
-			MethodName: "LoadModel",
-			Handler:    _Backend_LoadModel_Handler,
-		},
-		{
-			MethodName: "Embedding",
-			Handler:    _Backend_Embedding_Handler,
-		},
-		{
-			MethodName: "GenerateImage",
-			Handler:    _Backend_GenerateImage_Handler,
-		},
-		{
-			MethodName: "AudioTranscription",
-			Handler:    _Backend_AudioTranscription_Handler,
-		},
-		{
-			MethodName: "TTS",
-			Handler:    _Backend_TTS_Handler,
-		},
-		{
-			MethodName: "TokenizeString",
-			Handler:    _Backend_TokenizeString_Handler,
-		},
-		{
-			MethodName: "Status",
-			Handler:    _Backend_Status_Handler,
-		},
-	},
-	Streams: []grpc.StreamDesc{
-		{
-			StreamName:    "PredictStream",
-			Handler:       _Backend_PredictStream_Handler,
-			ServerStreams: true,
-		},
-	},
-	Metadata: "backend/backend.proto",
-}
diff --git a/backend/python/autogptq/Makefile b/backend/python/autogptq/Makefile
index dfae12c1..eb81f045 100644
--- a/backend/python/autogptq/Makefile
+++ b/backend/python/autogptq/Makefile
@@ -1,4 +1,13 @@
 .PHONY: autogptq
-autogptq:
+autogptq: protogen
 	$(MAKE) -C ../common-env/transformers
 
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/autogptq/backend_pb2.py b/backend/python/autogptq/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/autogptq/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/autogptq/backend_pb2_grpc.py b/backend/python/autogptq/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/autogptq/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/bark/Makefile b/backend/python/bark/Makefile
index 68f73b29..a16308f7 100644
--- a/backend/python/bark/Makefile
+++ b/backend/python/bark/Makefile
@@ -1,15 +1,25 @@
 .PHONY: ttsbark
-ttsbark:
+ttsbark: protogen
 	$(MAKE) -C ../common-env/transformers
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running bark..."
 	bash run.sh
 	@echo "bark run."
 
 .PHONY: test
-test:
+test: protogen
 	@echo "Testing bark..."
 	bash test.sh
 	@echo "bark tested."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/bark/backend_pb2.py b/backend/python/bark/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/bark/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/bark/backend_pb2_grpc.py b/backend/python/bark/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/bark/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/coqui/Makefile b/backend/python/coqui/Makefile
index e0ec9001..475804c9 100644
--- a/backend/python/coqui/Makefile
+++ b/backend/python/coqui/Makefile
@@ -1,15 +1,25 @@
 .PHONY: coqui
-coqui:
+coqui: protogen
 	$(MAKE) -C ../common-env/transformers
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running coqui..."
 	bash run.sh
 	@echo "coqui run."
 
 .PHONY: test
-test:
+test: protogen
 	@echo "Testing coqui..."
 	bash test.sh
 	@echo "coqui tested."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/coqui/backend_pb2.py b/backend/python/coqui/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/coqui/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/coqui/backend_pb2_grpc.py b/backend/python/coqui/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/coqui/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/diffusers/Makefile b/backend/python/diffusers/Makefile
index 40e1d1a7..c73efdd2 100644
--- a/backend/python/diffusers/Makefile
+++ b/backend/python/diffusers/Makefile
@@ -12,15 +12,25 @@ export SKIP_CONDA=1
 endif
 
 .PHONY: diffusers
-diffusers:
+diffusers: protogen
 	@echo "Installing $(CONDA_ENV_PATH)..."
 	bash install.sh $(CONDA_ENV_PATH)
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running diffusers..."
 	bash run.sh
 	@echo "Diffusers run."
 
-test:
+test: protogen
 	bash test.sh
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/diffusers/backend_pb2.py b/backend/python/diffusers/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/diffusers/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/diffusers/backend_pb2_grpc.py b/backend/python/diffusers/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/diffusers/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/exllama/Makefile b/backend/python/exllama/Makefile
index b51adf76..15623448 100644
--- a/backend/python/exllama/Makefile
+++ b/backend/python/exllama/Makefile
@@ -1,11 +1,21 @@
 export CONDA_ENV_PATH = "exllama.yml"
 
 .PHONY: exllama
-exllama:
+exllama: protogen
 	bash install.sh ${CONDA_ENV_PATH}
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running exllama..."
 	bash run.sh
 	@echo "exllama run."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/exllama/backend_pb2.py b/backend/python/exllama/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/exllama/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/exllama/backend_pb2_grpc.py b/backend/python/exllama/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/exllama/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/exllama2/Makefile b/backend/python/exllama2/Makefile
index 24158151..6d6776b7 100644
--- a/backend/python/exllama2/Makefile
+++ b/backend/python/exllama2/Makefile
@@ -1,10 +1,20 @@
 .PHONY: exllama2
-exllama2:
+exllama2: protogen
 	$(MAKE) -C ../common-env/transformers
 	bash install.sh
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running exllama2..."
 	bash run.sh
 	@echo "exllama2 run."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/exllama2/backend_pb2.py b/backend/python/exllama2/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/exllama2/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/exllama2/backend_pb2_grpc.py b/backend/python/exllama2/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/exllama2/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/mamba/Makefile b/backend/python/mamba/Makefile
index 3ff00346..ca18e609 100644
--- a/backend/python/mamba/Makefile
+++ b/backend/python/mamba/Makefile
@@ -1,16 +1,26 @@
 .PHONY: mamba
-mamba:
+mamba: protogen
 	$(MAKE) -C ../common-env/transformers
 	bash install.sh
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running mamba..."
 	bash run.sh
 	@echo "mamba run."
 
 .PHONY: test
-test:
+test: protogen
 	@echo "Testing mamba..."
 	bash test.sh
-	@echo "mamba tested."
\ No newline at end of file
+	@echo "mamba tested."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/mamba/backend_pb2.py b/backend/python/mamba/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/mamba/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/mamba/backend_pb2_grpc.py b/backend/python/mamba/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/mamba/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/petals/Makefile b/backend/python/petals/Makefile
index aa7778e1..0ed64a07 100644
--- a/backend/python/petals/Makefile
+++ b/backend/python/petals/Makefile
@@ -1,17 +1,27 @@
 .PHONY: petals
-petals:
+petals: protogen
 	@echo "Creating virtual environment..."
 	bash install.sh "petals.yml"
 	@echo "Virtual environment created."
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running petals..."
 	bash run.sh
 	@echo "petals run."
 
 .PHONY: test
-test:
+test: protogen
 	@echo "Testing petals..."
 	bash test.sh
 	@echo "petals tested."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/petals/backend_pb2.py b/backend/python/petals/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/petals/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/petals/backend_pb2_grpc.py b/backend/python/petals/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/petals/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/sentencetransformers/Makefile b/backend/python/sentencetransformers/Makefile
index 7dbde5cf..ac442897 100644
--- a/backend/python/sentencetransformers/Makefile
+++ b/backend/python/sentencetransformers/Makefile
@@ -1,17 +1,27 @@
 .PHONY: sentencetransformers
-sentencetransformers:
+sentencetransformers: protogen
 	$(MAKE) -C ../common-env/transformers
 
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running sentencetransformers..."
 	bash run.sh
 	@echo "sentencetransformers run."
 
 # It is not working well by using command line. It only6 works with IDE like VSCode.
 .PHONY: test
-test:
+test: protogen
 	@echo "Testing sentencetransformers..."
 	bash test.sh
 	@echo "sentencetransformers tested."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/sentencetransformers/backend_pb2.py b/backend/python/sentencetransformers/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/sentencetransformers/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/sentencetransformers/backend_pb2_grpc.py b/backend/python/sentencetransformers/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/sentencetransformers/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/transformers-musicgen/Makefile b/backend/python/transformers-musicgen/Makefile
index a2969d84..e28a356d 100644
--- a/backend/python/transformers-musicgen/Makefile
+++ b/backend/python/transformers-musicgen/Makefile
@@ -1,16 +1,25 @@
-
 .PHONY: transformers-musicgen
-transformers-musicgen:
+transformers-musicgen: protogen
 	$(MAKE) -C ../common-env/transformers
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running transformers..."
 	bash run.sh
 	@echo "transformers run."
 
 .PHONY: test
-test:
+test: protogen
 	@echo "Testing transformers..."
 	bash test.sh
 	@echo "transformers tested."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/transformers-musicgen/backend_pb2.py b/backend/python/transformers-musicgen/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/transformers-musicgen/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/transformers-musicgen/backend_pb2_grpc.py b/backend/python/transformers-musicgen/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/transformers-musicgen/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/transformers/Makefile b/backend/python/transformers/Makefile
index 4eeb9ad5..afe48405 100644
--- a/backend/python/transformers/Makefile
+++ b/backend/python/transformers/Makefile
@@ -1,16 +1,26 @@
 .PHONY: transformers
-transformers:
+transformers: protogen
 	$(MAKE) -C ../common-env/transformers
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running transformers..."
 	bash run.sh
 	@echo "transformers run."
 
 # It is not working well by using command line. It only6 works with IDE like VSCode.
 .PHONY: test
-test:
+test: protogen
 	@echo "Testing transformers..."
 	bash test.sh
 	@echo "transformers tested."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/transformers/backend_pb2.py b/backend/python/transformers/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/transformers/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/transformers/backend_pb2_grpc.py b/backend/python/transformers/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/transformers/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/vall-e-x/Makefile b/backend/python/vall-e-x/Makefile
index 8f34f559..d7a80e55 100644
--- a/backend/python/vall-e-x/Makefile
+++ b/backend/python/vall-e-x/Makefile
@@ -3,18 +3,28 @@ export SKIP_CONDA=1
 endif
 
 .PHONY: ttsvalle
-ttsvalle:
+ttsvalle: protogen
 	$(MAKE) -C ../common-env/transformers
 	bash install.sh
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running ttsvalle..."
 	bash run.sh
 	@echo "ttsvalle run."
 
 .PHONY: test
-test:
+test: protogen
 	@echo "Testing valle..."
 	bash test.sh
 	@echo "valle tested."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/vall-e-x/backend_pb2.py b/backend/python/vall-e-x/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/vall-e-x/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/vall-e-x/backend_pb2_grpc.py b/backend/python/vall-e-x/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/vall-e-x/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/backend/python/vllm/Makefile b/backend/python/vllm/Makefile
index 9ee5886d..3e1fdd77 100644
--- a/backend/python/vllm/Makefile
+++ b/backend/python/vllm/Makefile
@@ -1,15 +1,25 @@
 .PHONY: vllm
-vllm:
+vllm: protogen
 	$(MAKE) -C ../common-env/transformers
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running vllm..."
 	bash run.sh
 	@echo "vllm run."
 
 .PHONY: test
-test:
+test: protogen
 	@echo "Testing vllm..."
 	bash test.sh
-	@echo "vllm tested."
\ No newline at end of file
+	@echo "vllm tested."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
\ No newline at end of file
diff --git a/backend/python/vllm/backend_pb2.py b/backend/python/vllm/backend_pb2.py
deleted file mode 100644
index 24b6de3b..00000000
--- a/backend/python/vllm/backend_pb2.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# -*- coding: utf-8 -*-
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: backend.proto
-# Protobuf Python Version: 4.25.1
-"""Generated protocol buffer code."""
-from google.protobuf import descriptor as _descriptor
-from google.protobuf import descriptor_pool as _descriptor_pool
-from google.protobuf import symbol_database as _symbol_database
-from google.protobuf.internal import builder as _builder
-# @@protoc_insertion_point(imports)
-
-_sym_db = _symbol_database.Default()
-
-
-
-
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x1b\n\tStoresKey\x12\x0e\n\x06\x46loats\x18\x01 \x03(\x02\"\x1c\n\x0bStoresValue\x12\r\n\x05\x42ytes\x18\x01 \x01(\x0c\"Z\n\x10StoresSetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"7\n\x13StoresDeleteOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"4\n\x10StoresGetOptions\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\"Y\n\x0fStoresGetResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\"B\n\x11StoresFindOptions\x12\x1f\n\x03Key\x18\x01 \x01(\x0b\x32\x12.backend.StoresKey\x12\x0c\n\x04TopK\x18\x02 \x01(\x05\"p\n\x10StoresFindResult\x12 \n\x04Keys\x18\x01 \x03(\x0b\x32\x12.backend.StoresKey\x12$\n\x06Values\x18\x02 \x03(\x0b\x32\x14.backend.StoresValue\x12\x14\n\x0cSimilarities\x18\x03 \x03(\x02\"\x0f\n\rHealthMessage\"\xe8\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\x12\x0e\n\x06NDraft\x18) \x01(\x05\x12\x0e\n\x06Images\x18* \x03(\t\x12\x1c\n\x14UseTokenizerTemplate\x18+ \x01(\x08\x12\"\n\x08Messages\x18, \x03(\x0b\x32\x10.backend.Message\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xb0\x08\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\x12\x11\n\tModelFile\x18\x15 \x01(\t\x12\x0e\n\x06\x44\x65vice\x18\x16 \x01(\t\x12\x11\n\tUseTriton\x18\x17 \x01(\x08\x12\x15\n\rModelBaseName\x18\x18 \x01(\t\x12\x18\n\x10UseFastTokenizer\x18\x19 \x01(\x08\x12\x14\n\x0cPipelineType\x18\x1a \x01(\t\x12\x15\n\rSchedulerType\x18\x1b \x01(\t\x12\x0c\n\x04\x43UDA\x18\x1c \x01(\x08\x12\x10\n\x08\x43\x46GScale\x18\x1d \x01(\x02\x12\x0f\n\x07IMG2IMG\x18\x1e \x01(\x08\x12\x11\n\tCLIPModel\x18\x1f \x01(\t\x12\x15\n\rCLIPSubfolder\x18  \x01(\t\x12\x10\n\x08\x43LIPSkip\x18! \x01(\x05\x12\x12\n\nControlNet\x18\x30 \x01(\t\x12\x11\n\tTokenizer\x18\" \x01(\t\x12\x10\n\x08LoraBase\x18# \x01(\t\x12\x13\n\x0bLoraAdapter\x18$ \x01(\t\x12\x11\n\tLoraScale\x18* \x01(\x02\x12\x11\n\tNoMulMatQ\x18% \x01(\x08\x12\x12\n\nDraftModel\x18\' \x01(\t\x12\x11\n\tAudioPath\x18& \x01(\t\x12\x14\n\x0cQuantization\x18( \x01(\t\x12\x1c\n\x14GPUMemoryUtilization\x18\x32 \x01(\x02\x12\x17\n\x0fTrustRemoteCode\x18\x33 \x01(\x08\x12\x14\n\x0c\x45nforceEager\x18\x34 \x01(\x08\x12\x11\n\tSwapSpace\x18\x35 \x01(\x05\x12\x13\n\x0bMaxModelLen\x18\x36 \x01(\x05\x12\x0e\n\x06MMProj\x18) \x01(\t\x12\x13\n\x0bRopeScaling\x18+ \x01(\t\x12\x15\n\rYarnExtFactor\x18, \x01(\x02\x12\x16\n\x0eYarnAttnFactor\x18- \x01(\x02\x12\x14\n\x0cYarnBetaFast\x18. \x01(\x02\x12\x14\n\x0cYarnBetaSlow\x18/ \x01(\x02\x12\x0c\n\x04Type\x18\x31 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\xd7\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\x12\x0b\n\x03src\x18\t \x01(\t\x12\x18\n\x10\x45nableParameters\x18\n \x01(\t\x12\x10\n\x08\x43LIPSkip\x18\x0b \x01(\x05\"E\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t\x12\r\n\x05voice\x18\x04 \x01(\t\"6\n\x14TokenizationResponse\x12\x0e\n\x06length\x18\x01 \x01(\x05\x12\x0e\n\x06tokens\x18\x02 \x03(\x05\"\x8e\x01\n\x0fMemoryUsageData\x12\r\n\x05total\x18\x01 \x01(\x04\x12:\n\tbreakdown\x18\x02 \x03(\x0b\x32\'.backend.MemoryUsageData.BreakdownEntry\x1a\x30\n\x0e\x42reakdownEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x04:\x02\x38\x01\"\xad\x01\n\x0eStatusResponse\x12,\n\x05state\x18\x01 \x01(\x0e\x32\x1d.backend.StatusResponse.State\x12(\n\x06memory\x18\x02 \x01(\x0b\x32\x18.backend.MemoryUsageData\"C\n\x05State\x12\x11\n\rUNINITIALIZED\x10\x00\x12\x08\n\x04\x42USY\x10\x01\x12\t\n\x05READY\x10\x02\x12\x12\n\x05\x45RROR\x10\xff\xff\xff\xff\xff\xff\xff\xff\xff\x01\"(\n\x07Message\x12\x0c\n\x04role\x18\x01 \x01(\t\x12\x0f\n\x07\x63ontent\x18\x02 \x01(\t2\xfb\x06\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x12J\n\x0eTokenizeString\x12\x17.backend.PredictOptions\x1a\x1d.backend.TokenizationResponse\"\x00\x12;\n\x06Status\x12\x16.backend.HealthMessage\x1a\x17.backend.StatusResponse\"\x00\x12\x39\n\tStoresSet\x12\x19.backend.StoresSetOptions\x1a\x0f.backend.Result\"\x00\x12?\n\x0cStoresDelete\x12\x1c.backend.StoresDeleteOptions\x1a\x0f.backend.Result\"\x00\x12\x42\n\tStoresGet\x12\x19.backend.StoresGetOptions\x1a\x18.backend.StoresGetResult\"\x00\x12\x45\n\nStoresFind\x12\x1a.backend.StoresFindOptions\x1a\x19.backend.StoresFindResult\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
-
-_globals = globals()
-_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
-_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'backend_pb2', _globals)
-if _descriptor._USE_C_DESCRIPTORS == False:
-  _globals['DESCRIPTOR']._options = None
-  _globals['DESCRIPTOR']._serialized_options = b'\n\031io.skynet.localai.backendB\016LocalAIBackendP\001Z+github.com/go-skynet/LocalAI/pkg/grpc/proto'
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._options = None
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_options = b'8\001'
-  _globals['_STORESKEY']._serialized_start=26
-  _globals['_STORESKEY']._serialized_end=53
-  _globals['_STORESVALUE']._serialized_start=55
-  _globals['_STORESVALUE']._serialized_end=83
-  _globals['_STORESSETOPTIONS']._serialized_start=85
-  _globals['_STORESSETOPTIONS']._serialized_end=175
-  _globals['_STORESDELETEOPTIONS']._serialized_start=177
-  _globals['_STORESDELETEOPTIONS']._serialized_end=232
-  _globals['_STORESGETOPTIONS']._serialized_start=234
-  _globals['_STORESGETOPTIONS']._serialized_end=286
-  _globals['_STORESGETRESULT']._serialized_start=288
-  _globals['_STORESGETRESULT']._serialized_end=377
-  _globals['_STORESFINDOPTIONS']._serialized_start=379
-  _globals['_STORESFINDOPTIONS']._serialized_end=445
-  _globals['_STORESFINDRESULT']._serialized_start=447
-  _globals['_STORESFINDRESULT']._serialized_end=559
-  _globals['_HEALTHMESSAGE']._serialized_start=561
-  _globals['_HEALTHMESSAGE']._serialized_end=576
-  _globals['_PREDICTOPTIONS']._serialized_start=579
-  _globals['_PREDICTOPTIONS']._serialized_end=1451
-  _globals['_REPLY']._serialized_start=1453
-  _globals['_REPLY']._serialized_end=1477
-  _globals['_MODELOPTIONS']._serialized_start=1480
-  _globals['_MODELOPTIONS']._serialized_end=2552
-  _globals['_RESULT']._serialized_start=2554
-  _globals['_RESULT']._serialized_end=2596
-  _globals['_EMBEDDINGRESULT']._serialized_start=2598
-  _globals['_EMBEDDINGRESULT']._serialized_end=2635
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=2637
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=2704
-  _globals['_TRANSCRIPTRESULT']._serialized_start=2706
-  _globals['_TRANSCRIPTRESULT']._serialized_end=2784
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=2786
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=2875
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=2878
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=3093
-  _globals['_TTSREQUEST']._serialized_start=3095
-  _globals['_TTSREQUEST']._serialized_end=3164
-  _globals['_TOKENIZATIONRESPONSE']._serialized_start=3166
-  _globals['_TOKENIZATIONRESPONSE']._serialized_end=3220
-  _globals['_MEMORYUSAGEDATA']._serialized_start=3223
-  _globals['_MEMORYUSAGEDATA']._serialized_end=3365
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_start=3317
-  _globals['_MEMORYUSAGEDATA_BREAKDOWNENTRY']._serialized_end=3365
-  _globals['_STATUSRESPONSE']._serialized_start=3368
-  _globals['_STATUSRESPONSE']._serialized_end=3541
-  _globals['_STATUSRESPONSE_STATE']._serialized_start=3474
-  _globals['_STATUSRESPONSE_STATE']._serialized_end=3541
-  _globals['_MESSAGE']._serialized_start=3543
-  _globals['_MESSAGE']._serialized_end=3583
-  _globals['_BACKEND']._serialized_start=3586
-  _globals['_BACKEND']._serialized_end=4477
-# @@protoc_insertion_point(module_scope)
diff --git a/backend/python/vllm/backend_pb2_grpc.py b/backend/python/vllm/backend_pb2_grpc.py
deleted file mode 100644
index e06fccf3..00000000
--- a/backend/python/vllm/backend_pb2_grpc.py
+++ /dev/null
@@ -1,495 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-        self.StoresSet = channel.unary_unary(
-                '/backend.Backend/StoresSet',
-                request_serializer=backend__pb2.StoresSetOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresDelete = channel.unary_unary(
-                '/backend.Backend/StoresDelete',
-                request_serializer=backend__pb2.StoresDeleteOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.StoresGet = channel.unary_unary(
-                '/backend.Backend/StoresGet',
-                request_serializer=backend__pb2.StoresGetOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresGetResult.FromString,
-                )
-        self.StoresFind = channel.unary_unary(
-                '/backend.Backend/StoresFind',
-                request_serializer=backend__pb2.StoresFindOptions.SerializeToString,
-                response_deserializer=backend__pb2.StoresFindResult.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresSet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresDelete(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresGet(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def StoresFind(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-            'StoresSet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresSet,
-                    request_deserializer=backend__pb2.StoresSetOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresDelete': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresDelete,
-                    request_deserializer=backend__pb2.StoresDeleteOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'StoresGet': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresGet,
-                    request_deserializer=backend__pb2.StoresGetOptions.FromString,
-                    response_serializer=backend__pb2.StoresGetResult.SerializeToString,
-            ),
-            'StoresFind': grpc.unary_unary_rpc_method_handler(
-                    servicer.StoresFind,
-                    request_deserializer=backend__pb2.StoresFindOptions.FromString,
-                    response_serializer=backend__pb2.StoresFindResult.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresSet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresSet',
-            backend__pb2.StoresSetOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresDelete(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresDelete',
-            backend__pb2.StoresDeleteOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresGet(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresGet',
-            backend__pb2.StoresGetOptions.SerializeToString,
-            backend__pb2.StoresGetResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def StoresFind(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/StoresFind',
-            backend__pb2.StoresFindOptions.SerializeToString,
-            backend__pb2.StoresFindResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
diff --git a/docs/content/docs/getting-started/build.md b/docs/content/docs/getting-started/build.md
index 8ceaf1f5..a4db135e 100644
--- a/docs/content/docs/getting-started/build.md
+++ b/docs/content/docs/getting-started/build.md
@@ -36,14 +36,28 @@ To install the dependencies follow the instructions below:
 Install `xcode` from the App Store
 
 ```bash
-brew install abseil cmake go grpc protobuf wget
+brew install abseil cmake go grpc protobuf protoc-gen-go protoc-gen-go-grpc python wget
+```
+
+After installing the above dependencies, you need to install grpcio-tools from PyPI. You could do this via a pip --user install or a virtualenv.
+
+```bash
+pip install --user grpcio-tools
 ```
 
 {{% /tab %}}
 {{% tab tabName="Debian" %}}
 
 ```bash
-apt install golang protobuf-compiler-grpc libgrpc-dev make cmake
+apt install cmake golang libgrpc-dev make protobuf-compiler-grpc python3-grpc-tools
+```
+
+After you have golang installed and working, you can install the required binaries for compiling the golang protobuf components via the following commands
+
+```bash
+go install google.golang.org/protobuf/cmd/protoc-gen-go@latest
+go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@latest
+
 ```
 
 {{% /tab %}}
diff --git a/pkg/grpc/proto/backend.pb.go b/pkg/grpc/proto/backend.pb.go
deleted file mode 100644
index e9afe196..00000000
--- a/pkg/grpc/proto/backend.pb.go
+++ /dev/null
@@ -1,2934 +0,0 @@
-// Code generated by protoc-gen-go. DO NOT EDIT.
-// versions:
-// 	protoc-gen-go v1.26.0
-// 	protoc        v5.26.1
-// source: backend.proto
-
-package proto
-
-import (
-	protoreflect "google.golang.org/protobuf/reflect/protoreflect"
-	protoimpl "google.golang.org/protobuf/runtime/protoimpl"
-	reflect "reflect"
-	sync "sync"
-)
-
-const (
-	// Verify that this generated code is sufficiently up-to-date.
-	_ = protoimpl.EnforceVersion(20 - protoimpl.MinVersion)
-	// Verify that runtime/protoimpl is sufficiently up-to-date.
-	_ = protoimpl.EnforceVersion(protoimpl.MaxVersion - 20)
-)
-
-type StatusResponse_State int32
-
-const (
-	StatusResponse_UNINITIALIZED StatusResponse_State = 0
-	StatusResponse_BUSY          StatusResponse_State = 1
-	StatusResponse_READY         StatusResponse_State = 2
-	StatusResponse_ERROR         StatusResponse_State = -1
-)
-
-// Enum value maps for StatusResponse_State.
-var (
-	StatusResponse_State_name = map[int32]string{
-		0:  "UNINITIALIZED",
-		1:  "BUSY",
-		2:  "READY",
-		-1: "ERROR",
-	}
-	StatusResponse_State_value = map[string]int32{
-		"UNINITIALIZED": 0,
-		"BUSY":          1,
-		"READY":         2,
-		"ERROR":         -1,
-	}
-)
-
-func (x StatusResponse_State) Enum() *StatusResponse_State {
-	p := new(StatusResponse_State)
-	*p = x
-	return p
-}
-
-func (x StatusResponse_State) String() string {
-	return protoimpl.X.EnumStringOf(x.Descriptor(), protoreflect.EnumNumber(x))
-}
-
-func (StatusResponse_State) Descriptor() protoreflect.EnumDescriptor {
-	return file_backend_proto_enumTypes[0].Descriptor()
-}
-
-func (StatusResponse_State) Type() protoreflect.EnumType {
-	return &file_backend_proto_enumTypes[0]
-}
-
-func (x StatusResponse_State) Number() protoreflect.EnumNumber {
-	return protoreflect.EnumNumber(x)
-}
-
-// Deprecated: Use StatusResponse_State.Descriptor instead.
-func (StatusResponse_State) EnumDescriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{21, 0}
-}
-
-type StoresKey struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Floats []float32 `protobuf:"fixed32,1,rep,packed,name=Floats,proto3" json:"Floats,omitempty"`
-}
-
-func (x *StoresKey) Reset() {
-	*x = StoresKey{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[0]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *StoresKey) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*StoresKey) ProtoMessage() {}
-
-func (x *StoresKey) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[0]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use StoresKey.ProtoReflect.Descriptor instead.
-func (*StoresKey) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{0}
-}
-
-func (x *StoresKey) GetFloats() []float32 {
-	if x != nil {
-		return x.Floats
-	}
-	return nil
-}
-
-type StoresValue struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Bytes []byte `protobuf:"bytes,1,opt,name=Bytes,proto3" json:"Bytes,omitempty"`
-}
-
-func (x *StoresValue) Reset() {
-	*x = StoresValue{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[1]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *StoresValue) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*StoresValue) ProtoMessage() {}
-
-func (x *StoresValue) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[1]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use StoresValue.ProtoReflect.Descriptor instead.
-func (*StoresValue) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{1}
-}
-
-func (x *StoresValue) GetBytes() []byte {
-	if x != nil {
-		return x.Bytes
-	}
-	return nil
-}
-
-type StoresSetOptions struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Keys   []*StoresKey   `protobuf:"bytes,1,rep,name=Keys,proto3" json:"Keys,omitempty"`
-	Values []*StoresValue `protobuf:"bytes,2,rep,name=Values,proto3" json:"Values,omitempty"`
-}
-
-func (x *StoresSetOptions) Reset() {
-	*x = StoresSetOptions{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[2]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *StoresSetOptions) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*StoresSetOptions) ProtoMessage() {}
-
-func (x *StoresSetOptions) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[2]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use StoresSetOptions.ProtoReflect.Descriptor instead.
-func (*StoresSetOptions) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{2}
-}
-
-func (x *StoresSetOptions) GetKeys() []*StoresKey {
-	if x != nil {
-		return x.Keys
-	}
-	return nil
-}
-
-func (x *StoresSetOptions) GetValues() []*StoresValue {
-	if x != nil {
-		return x.Values
-	}
-	return nil
-}
-
-type StoresDeleteOptions struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Keys []*StoresKey `protobuf:"bytes,1,rep,name=Keys,proto3" json:"Keys,omitempty"`
-}
-
-func (x *StoresDeleteOptions) Reset() {
-	*x = StoresDeleteOptions{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[3]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *StoresDeleteOptions) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*StoresDeleteOptions) ProtoMessage() {}
-
-func (x *StoresDeleteOptions) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[3]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use StoresDeleteOptions.ProtoReflect.Descriptor instead.
-func (*StoresDeleteOptions) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{3}
-}
-
-func (x *StoresDeleteOptions) GetKeys() []*StoresKey {
-	if x != nil {
-		return x.Keys
-	}
-	return nil
-}
-
-type StoresGetOptions struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Keys []*StoresKey `protobuf:"bytes,1,rep,name=Keys,proto3" json:"Keys,omitempty"`
-}
-
-func (x *StoresGetOptions) Reset() {
-	*x = StoresGetOptions{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[4]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *StoresGetOptions) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*StoresGetOptions) ProtoMessage() {}
-
-func (x *StoresGetOptions) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[4]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use StoresGetOptions.ProtoReflect.Descriptor instead.
-func (*StoresGetOptions) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{4}
-}
-
-func (x *StoresGetOptions) GetKeys() []*StoresKey {
-	if x != nil {
-		return x.Keys
-	}
-	return nil
-}
-
-type StoresGetResult struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Keys   []*StoresKey   `protobuf:"bytes,1,rep,name=Keys,proto3" json:"Keys,omitempty"`
-	Values []*StoresValue `protobuf:"bytes,2,rep,name=Values,proto3" json:"Values,omitempty"`
-}
-
-func (x *StoresGetResult) Reset() {
-	*x = StoresGetResult{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[5]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *StoresGetResult) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*StoresGetResult) ProtoMessage() {}
-
-func (x *StoresGetResult) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[5]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use StoresGetResult.ProtoReflect.Descriptor instead.
-func (*StoresGetResult) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{5}
-}
-
-func (x *StoresGetResult) GetKeys() []*StoresKey {
-	if x != nil {
-		return x.Keys
-	}
-	return nil
-}
-
-func (x *StoresGetResult) GetValues() []*StoresValue {
-	if x != nil {
-		return x.Values
-	}
-	return nil
-}
-
-type StoresFindOptions struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Key  *StoresKey `protobuf:"bytes,1,opt,name=Key,proto3" json:"Key,omitempty"`
-	TopK int32      `protobuf:"varint,2,opt,name=TopK,proto3" json:"TopK,omitempty"`
-}
-
-func (x *StoresFindOptions) Reset() {
-	*x = StoresFindOptions{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[6]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *StoresFindOptions) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*StoresFindOptions) ProtoMessage() {}
-
-func (x *StoresFindOptions) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[6]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use StoresFindOptions.ProtoReflect.Descriptor instead.
-func (*StoresFindOptions) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{6}
-}
-
-func (x *StoresFindOptions) GetKey() *StoresKey {
-	if x != nil {
-		return x.Key
-	}
-	return nil
-}
-
-func (x *StoresFindOptions) GetTopK() int32 {
-	if x != nil {
-		return x.TopK
-	}
-	return 0
-}
-
-type StoresFindResult struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Keys         []*StoresKey   `protobuf:"bytes,1,rep,name=Keys,proto3" json:"Keys,omitempty"`
-	Values       []*StoresValue `protobuf:"bytes,2,rep,name=Values,proto3" json:"Values,omitempty"`
-	Similarities []float32      `protobuf:"fixed32,3,rep,packed,name=Similarities,proto3" json:"Similarities,omitempty"`
-}
-
-func (x *StoresFindResult) Reset() {
-	*x = StoresFindResult{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[7]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *StoresFindResult) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*StoresFindResult) ProtoMessage() {}
-
-func (x *StoresFindResult) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[7]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use StoresFindResult.ProtoReflect.Descriptor instead.
-func (*StoresFindResult) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{7}
-}
-
-func (x *StoresFindResult) GetKeys() []*StoresKey {
-	if x != nil {
-		return x.Keys
-	}
-	return nil
-}
-
-func (x *StoresFindResult) GetValues() []*StoresValue {
-	if x != nil {
-		return x.Values
-	}
-	return nil
-}
-
-func (x *StoresFindResult) GetSimilarities() []float32 {
-	if x != nil {
-		return x.Similarities
-	}
-	return nil
-}
-
-type HealthMessage struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-}
-
-func (x *HealthMessage) Reset() {
-	*x = HealthMessage{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[8]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *HealthMessage) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*HealthMessage) ProtoMessage() {}
-
-func (x *HealthMessage) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[8]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use HealthMessage.ProtoReflect.Descriptor instead.
-func (*HealthMessage) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{8}
-}
-
-// The request message containing the user's name.
-type PredictOptions struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Prompt               string     `protobuf:"bytes,1,opt,name=Prompt,proto3" json:"Prompt,omitempty"`
-	Seed                 int32      `protobuf:"varint,2,opt,name=Seed,proto3" json:"Seed,omitempty"`
-	Threads              int32      `protobuf:"varint,3,opt,name=Threads,proto3" json:"Threads,omitempty"`
-	Tokens               int32      `protobuf:"varint,4,opt,name=Tokens,proto3" json:"Tokens,omitempty"`
-	TopK                 int32      `protobuf:"varint,5,opt,name=TopK,proto3" json:"TopK,omitempty"`
-	Repeat               int32      `protobuf:"varint,6,opt,name=Repeat,proto3" json:"Repeat,omitempty"`
-	Batch                int32      `protobuf:"varint,7,opt,name=Batch,proto3" json:"Batch,omitempty"`
-	NKeep                int32      `protobuf:"varint,8,opt,name=NKeep,proto3" json:"NKeep,omitempty"`
-	Temperature          float32    `protobuf:"fixed32,9,opt,name=Temperature,proto3" json:"Temperature,omitempty"`
-	Penalty              float32    `protobuf:"fixed32,10,opt,name=Penalty,proto3" json:"Penalty,omitempty"`
-	F16KV                bool       `protobuf:"varint,11,opt,name=F16KV,proto3" json:"F16KV,omitempty"`
-	DebugMode            bool       `protobuf:"varint,12,opt,name=DebugMode,proto3" json:"DebugMode,omitempty"`
-	StopPrompts          []string   `protobuf:"bytes,13,rep,name=StopPrompts,proto3" json:"StopPrompts,omitempty"`
-	IgnoreEOS            bool       `protobuf:"varint,14,opt,name=IgnoreEOS,proto3" json:"IgnoreEOS,omitempty"`
-	TailFreeSamplingZ    float32    `protobuf:"fixed32,15,opt,name=TailFreeSamplingZ,proto3" json:"TailFreeSamplingZ,omitempty"`
-	TypicalP             float32    `protobuf:"fixed32,16,opt,name=TypicalP,proto3" json:"TypicalP,omitempty"`
-	FrequencyPenalty     float32    `protobuf:"fixed32,17,opt,name=FrequencyPenalty,proto3" json:"FrequencyPenalty,omitempty"`
-	PresencePenalty      float32    `protobuf:"fixed32,18,opt,name=PresencePenalty,proto3" json:"PresencePenalty,omitempty"`
-	Mirostat             int32      `protobuf:"varint,19,opt,name=Mirostat,proto3" json:"Mirostat,omitempty"`
-	MirostatETA          float32    `protobuf:"fixed32,20,opt,name=MirostatETA,proto3" json:"MirostatETA,omitempty"`
-	MirostatTAU          float32    `protobuf:"fixed32,21,opt,name=MirostatTAU,proto3" json:"MirostatTAU,omitempty"`
-	PenalizeNL           bool       `protobuf:"varint,22,opt,name=PenalizeNL,proto3" json:"PenalizeNL,omitempty"`
-	LogitBias            string     `protobuf:"bytes,23,opt,name=LogitBias,proto3" json:"LogitBias,omitempty"`
-	MLock                bool       `protobuf:"varint,25,opt,name=MLock,proto3" json:"MLock,omitempty"`
-	MMap                 bool       `protobuf:"varint,26,opt,name=MMap,proto3" json:"MMap,omitempty"`
-	PromptCacheAll       bool       `protobuf:"varint,27,opt,name=PromptCacheAll,proto3" json:"PromptCacheAll,omitempty"`
-	PromptCacheRO        bool       `protobuf:"varint,28,opt,name=PromptCacheRO,proto3" json:"PromptCacheRO,omitempty"`
-	Grammar              string     `protobuf:"bytes,29,opt,name=Grammar,proto3" json:"Grammar,omitempty"`
-	MainGPU              string     `protobuf:"bytes,30,opt,name=MainGPU,proto3" json:"MainGPU,omitempty"`
-	TensorSplit          string     `protobuf:"bytes,31,opt,name=TensorSplit,proto3" json:"TensorSplit,omitempty"`
-	TopP                 float32    `protobuf:"fixed32,32,opt,name=TopP,proto3" json:"TopP,omitempty"`
-	PromptCachePath      string     `protobuf:"bytes,33,opt,name=PromptCachePath,proto3" json:"PromptCachePath,omitempty"`
-	Debug                bool       `protobuf:"varint,34,opt,name=Debug,proto3" json:"Debug,omitempty"`
-	EmbeddingTokens      []int32    `protobuf:"varint,35,rep,packed,name=EmbeddingTokens,proto3" json:"EmbeddingTokens,omitempty"`
-	Embeddings           string     `protobuf:"bytes,36,opt,name=Embeddings,proto3" json:"Embeddings,omitempty"`
-	RopeFreqBase         float32    `protobuf:"fixed32,37,opt,name=RopeFreqBase,proto3" json:"RopeFreqBase,omitempty"`
-	RopeFreqScale        float32    `protobuf:"fixed32,38,opt,name=RopeFreqScale,proto3" json:"RopeFreqScale,omitempty"`
-	NegativePromptScale  float32    `protobuf:"fixed32,39,opt,name=NegativePromptScale,proto3" json:"NegativePromptScale,omitempty"`
-	NegativePrompt       string     `protobuf:"bytes,40,opt,name=NegativePrompt,proto3" json:"NegativePrompt,omitempty"`
-	NDraft               int32      `protobuf:"varint,41,opt,name=NDraft,proto3" json:"NDraft,omitempty"`
-	Images               []string   `protobuf:"bytes,42,rep,name=Images,proto3" json:"Images,omitempty"`
-	UseTokenizerTemplate bool       `protobuf:"varint,43,opt,name=UseTokenizerTemplate,proto3" json:"UseTokenizerTemplate,omitempty"`
-	Messages             []*Message `protobuf:"bytes,44,rep,name=Messages,proto3" json:"Messages,omitempty"`
-}
-
-func (x *PredictOptions) Reset() {
-	*x = PredictOptions{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[9]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *PredictOptions) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*PredictOptions) ProtoMessage() {}
-
-func (x *PredictOptions) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[9]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use PredictOptions.ProtoReflect.Descriptor instead.
-func (*PredictOptions) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{9}
-}
-
-func (x *PredictOptions) GetPrompt() string {
-	if x != nil {
-		return x.Prompt
-	}
-	return ""
-}
-
-func (x *PredictOptions) GetSeed() int32 {
-	if x != nil {
-		return x.Seed
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetThreads() int32 {
-	if x != nil {
-		return x.Threads
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetTokens() int32 {
-	if x != nil {
-		return x.Tokens
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetTopK() int32 {
-	if x != nil {
-		return x.TopK
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetRepeat() int32 {
-	if x != nil {
-		return x.Repeat
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetBatch() int32 {
-	if x != nil {
-		return x.Batch
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetNKeep() int32 {
-	if x != nil {
-		return x.NKeep
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetTemperature() float32 {
-	if x != nil {
-		return x.Temperature
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetPenalty() float32 {
-	if x != nil {
-		return x.Penalty
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetF16KV() bool {
-	if x != nil {
-		return x.F16KV
-	}
-	return false
-}
-
-func (x *PredictOptions) GetDebugMode() bool {
-	if x != nil {
-		return x.DebugMode
-	}
-	return false
-}
-
-func (x *PredictOptions) GetStopPrompts() []string {
-	if x != nil {
-		return x.StopPrompts
-	}
-	return nil
-}
-
-func (x *PredictOptions) GetIgnoreEOS() bool {
-	if x != nil {
-		return x.IgnoreEOS
-	}
-	return false
-}
-
-func (x *PredictOptions) GetTailFreeSamplingZ() float32 {
-	if x != nil {
-		return x.TailFreeSamplingZ
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetTypicalP() float32 {
-	if x != nil {
-		return x.TypicalP
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetFrequencyPenalty() float32 {
-	if x != nil {
-		return x.FrequencyPenalty
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetPresencePenalty() float32 {
-	if x != nil {
-		return x.PresencePenalty
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetMirostat() int32 {
-	if x != nil {
-		return x.Mirostat
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetMirostatETA() float32 {
-	if x != nil {
-		return x.MirostatETA
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetMirostatTAU() float32 {
-	if x != nil {
-		return x.MirostatTAU
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetPenalizeNL() bool {
-	if x != nil {
-		return x.PenalizeNL
-	}
-	return false
-}
-
-func (x *PredictOptions) GetLogitBias() string {
-	if x != nil {
-		return x.LogitBias
-	}
-	return ""
-}
-
-func (x *PredictOptions) GetMLock() bool {
-	if x != nil {
-		return x.MLock
-	}
-	return false
-}
-
-func (x *PredictOptions) GetMMap() bool {
-	if x != nil {
-		return x.MMap
-	}
-	return false
-}
-
-func (x *PredictOptions) GetPromptCacheAll() bool {
-	if x != nil {
-		return x.PromptCacheAll
-	}
-	return false
-}
-
-func (x *PredictOptions) GetPromptCacheRO() bool {
-	if x != nil {
-		return x.PromptCacheRO
-	}
-	return false
-}
-
-func (x *PredictOptions) GetGrammar() string {
-	if x != nil {
-		return x.Grammar
-	}
-	return ""
-}
-
-func (x *PredictOptions) GetMainGPU() string {
-	if x != nil {
-		return x.MainGPU
-	}
-	return ""
-}
-
-func (x *PredictOptions) GetTensorSplit() string {
-	if x != nil {
-		return x.TensorSplit
-	}
-	return ""
-}
-
-func (x *PredictOptions) GetTopP() float32 {
-	if x != nil {
-		return x.TopP
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetPromptCachePath() string {
-	if x != nil {
-		return x.PromptCachePath
-	}
-	return ""
-}
-
-func (x *PredictOptions) GetDebug() bool {
-	if x != nil {
-		return x.Debug
-	}
-	return false
-}
-
-func (x *PredictOptions) GetEmbeddingTokens() []int32 {
-	if x != nil {
-		return x.EmbeddingTokens
-	}
-	return nil
-}
-
-func (x *PredictOptions) GetEmbeddings() string {
-	if x != nil {
-		return x.Embeddings
-	}
-	return ""
-}
-
-func (x *PredictOptions) GetRopeFreqBase() float32 {
-	if x != nil {
-		return x.RopeFreqBase
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetRopeFreqScale() float32 {
-	if x != nil {
-		return x.RopeFreqScale
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetNegativePromptScale() float32 {
-	if x != nil {
-		return x.NegativePromptScale
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetNegativePrompt() string {
-	if x != nil {
-		return x.NegativePrompt
-	}
-	return ""
-}
-
-func (x *PredictOptions) GetNDraft() int32 {
-	if x != nil {
-		return x.NDraft
-	}
-	return 0
-}
-
-func (x *PredictOptions) GetImages() []string {
-	if x != nil {
-		return x.Images
-	}
-	return nil
-}
-
-func (x *PredictOptions) GetUseTokenizerTemplate() bool {
-	if x != nil {
-		return x.UseTokenizerTemplate
-	}
-	return false
-}
-
-func (x *PredictOptions) GetMessages() []*Message {
-	if x != nil {
-		return x.Messages
-	}
-	return nil
-}
-
-// The response message containing the result
-type Reply struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Message []byte `protobuf:"bytes,1,opt,name=message,proto3" json:"message,omitempty"`
-}
-
-func (x *Reply) Reset() {
-	*x = Reply{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[10]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *Reply) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*Reply) ProtoMessage() {}
-
-func (x *Reply) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[10]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use Reply.ProtoReflect.Descriptor instead.
-func (*Reply) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{10}
-}
-
-func (x *Reply) GetMessage() []byte {
-	if x != nil {
-		return x.Message
-	}
-	return nil
-}
-
-type ModelOptions struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Model             string  `protobuf:"bytes,1,opt,name=Model,proto3" json:"Model,omitempty"`
-	ContextSize       int32   `protobuf:"varint,2,opt,name=ContextSize,proto3" json:"ContextSize,omitempty"`
-	Seed              int32   `protobuf:"varint,3,opt,name=Seed,proto3" json:"Seed,omitempty"`
-	NBatch            int32   `protobuf:"varint,4,opt,name=NBatch,proto3" json:"NBatch,omitempty"`
-	F16Memory         bool    `protobuf:"varint,5,opt,name=F16Memory,proto3" json:"F16Memory,omitempty"`
-	MLock             bool    `protobuf:"varint,6,opt,name=MLock,proto3" json:"MLock,omitempty"`
-	MMap              bool    `protobuf:"varint,7,opt,name=MMap,proto3" json:"MMap,omitempty"`
-	VocabOnly         bool    `protobuf:"varint,8,opt,name=VocabOnly,proto3" json:"VocabOnly,omitempty"`
-	LowVRAM           bool    `protobuf:"varint,9,opt,name=LowVRAM,proto3" json:"LowVRAM,omitempty"`
-	Embeddings        bool    `protobuf:"varint,10,opt,name=Embeddings,proto3" json:"Embeddings,omitempty"`
-	NUMA              bool    `protobuf:"varint,11,opt,name=NUMA,proto3" json:"NUMA,omitempty"`
-	NGPULayers        int32   `protobuf:"varint,12,opt,name=NGPULayers,proto3" json:"NGPULayers,omitempty"`
-	MainGPU           string  `protobuf:"bytes,13,opt,name=MainGPU,proto3" json:"MainGPU,omitempty"`
-	TensorSplit       string  `protobuf:"bytes,14,opt,name=TensorSplit,proto3" json:"TensorSplit,omitempty"`
-	Threads           int32   `protobuf:"varint,15,opt,name=Threads,proto3" json:"Threads,omitempty"`
-	LibrarySearchPath string  `protobuf:"bytes,16,opt,name=LibrarySearchPath,proto3" json:"LibrarySearchPath,omitempty"`
-	RopeFreqBase      float32 `protobuf:"fixed32,17,opt,name=RopeFreqBase,proto3" json:"RopeFreqBase,omitempty"`
-	RopeFreqScale     float32 `protobuf:"fixed32,18,opt,name=RopeFreqScale,proto3" json:"RopeFreqScale,omitempty"`
-	RMSNormEps        float32 `protobuf:"fixed32,19,opt,name=RMSNormEps,proto3" json:"RMSNormEps,omitempty"`
-	NGQA              int32   `protobuf:"varint,20,opt,name=NGQA,proto3" json:"NGQA,omitempty"`
-	ModelFile         string  `protobuf:"bytes,21,opt,name=ModelFile,proto3" json:"ModelFile,omitempty"`
-	// AutoGPTQ
-	Device           string `protobuf:"bytes,22,opt,name=Device,proto3" json:"Device,omitempty"`
-	UseTriton        bool   `protobuf:"varint,23,opt,name=UseTriton,proto3" json:"UseTriton,omitempty"`
-	ModelBaseName    string `protobuf:"bytes,24,opt,name=ModelBaseName,proto3" json:"ModelBaseName,omitempty"`
-	UseFastTokenizer bool   `protobuf:"varint,25,opt,name=UseFastTokenizer,proto3" json:"UseFastTokenizer,omitempty"`
-	// Diffusers
-	PipelineType  string  `protobuf:"bytes,26,opt,name=PipelineType,proto3" json:"PipelineType,omitempty"`
-	SchedulerType string  `protobuf:"bytes,27,opt,name=SchedulerType,proto3" json:"SchedulerType,omitempty"`
-	CUDA          bool    `protobuf:"varint,28,opt,name=CUDA,proto3" json:"CUDA,omitempty"`
-	CFGScale      float32 `protobuf:"fixed32,29,opt,name=CFGScale,proto3" json:"CFGScale,omitempty"`
-	IMG2IMG       bool    `protobuf:"varint,30,opt,name=IMG2IMG,proto3" json:"IMG2IMG,omitempty"`
-	CLIPModel     string  `protobuf:"bytes,31,opt,name=CLIPModel,proto3" json:"CLIPModel,omitempty"`
-	CLIPSubfolder string  `protobuf:"bytes,32,opt,name=CLIPSubfolder,proto3" json:"CLIPSubfolder,omitempty"`
-	CLIPSkip      int32   `protobuf:"varint,33,opt,name=CLIPSkip,proto3" json:"CLIPSkip,omitempty"`
-	ControlNet    string  `protobuf:"bytes,48,opt,name=ControlNet,proto3" json:"ControlNet,omitempty"`
-	Tokenizer     string  `protobuf:"bytes,34,opt,name=Tokenizer,proto3" json:"Tokenizer,omitempty"`
-	// LLM (llama.cpp)
-	LoraBase    string  `protobuf:"bytes,35,opt,name=LoraBase,proto3" json:"LoraBase,omitempty"`
-	LoraAdapter string  `protobuf:"bytes,36,opt,name=LoraAdapter,proto3" json:"LoraAdapter,omitempty"`
-	LoraScale   float32 `protobuf:"fixed32,42,opt,name=LoraScale,proto3" json:"LoraScale,omitempty"`
-	NoMulMatQ   bool    `protobuf:"varint,37,opt,name=NoMulMatQ,proto3" json:"NoMulMatQ,omitempty"`
-	DraftModel  string  `protobuf:"bytes,39,opt,name=DraftModel,proto3" json:"DraftModel,omitempty"`
-	AudioPath   string  `protobuf:"bytes,38,opt,name=AudioPath,proto3" json:"AudioPath,omitempty"`
-	// vllm
-	Quantization         string  `protobuf:"bytes,40,opt,name=Quantization,proto3" json:"Quantization,omitempty"`
-	GPUMemoryUtilization float32 `protobuf:"fixed32,50,opt,name=GPUMemoryUtilization,proto3" json:"GPUMemoryUtilization,omitempty"`
-	TrustRemoteCode      bool    `protobuf:"varint,51,opt,name=TrustRemoteCode,proto3" json:"TrustRemoteCode,omitempty"`
-	EnforceEager         bool    `protobuf:"varint,52,opt,name=EnforceEager,proto3" json:"EnforceEager,omitempty"`
-	SwapSpace            int32   `protobuf:"varint,53,opt,name=SwapSpace,proto3" json:"SwapSpace,omitempty"`
-	MaxModelLen          int32   `protobuf:"varint,54,opt,name=MaxModelLen,proto3" json:"MaxModelLen,omitempty"`
-	MMProj               string  `protobuf:"bytes,41,opt,name=MMProj,proto3" json:"MMProj,omitempty"`
-	RopeScaling          string  `protobuf:"bytes,43,opt,name=RopeScaling,proto3" json:"RopeScaling,omitempty"`
-	YarnExtFactor        float32 `protobuf:"fixed32,44,opt,name=YarnExtFactor,proto3" json:"YarnExtFactor,omitempty"`
-	YarnAttnFactor       float32 `protobuf:"fixed32,45,opt,name=YarnAttnFactor,proto3" json:"YarnAttnFactor,omitempty"`
-	YarnBetaFast         float32 `protobuf:"fixed32,46,opt,name=YarnBetaFast,proto3" json:"YarnBetaFast,omitempty"`
-	YarnBetaSlow         float32 `protobuf:"fixed32,47,opt,name=YarnBetaSlow,proto3" json:"YarnBetaSlow,omitempty"`
-	Type                 string  `protobuf:"bytes,49,opt,name=Type,proto3" json:"Type,omitempty"`
-}
-
-func (x *ModelOptions) Reset() {
-	*x = ModelOptions{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[11]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *ModelOptions) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*ModelOptions) ProtoMessage() {}
-
-func (x *ModelOptions) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[11]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use ModelOptions.ProtoReflect.Descriptor instead.
-func (*ModelOptions) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{11}
-}
-
-func (x *ModelOptions) GetModel() string {
-	if x != nil {
-		return x.Model
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetContextSize() int32 {
-	if x != nil {
-		return x.ContextSize
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetSeed() int32 {
-	if x != nil {
-		return x.Seed
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetNBatch() int32 {
-	if x != nil {
-		return x.NBatch
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetF16Memory() bool {
-	if x != nil {
-		return x.F16Memory
-	}
-	return false
-}
-
-func (x *ModelOptions) GetMLock() bool {
-	if x != nil {
-		return x.MLock
-	}
-	return false
-}
-
-func (x *ModelOptions) GetMMap() bool {
-	if x != nil {
-		return x.MMap
-	}
-	return false
-}
-
-func (x *ModelOptions) GetVocabOnly() bool {
-	if x != nil {
-		return x.VocabOnly
-	}
-	return false
-}
-
-func (x *ModelOptions) GetLowVRAM() bool {
-	if x != nil {
-		return x.LowVRAM
-	}
-	return false
-}
-
-func (x *ModelOptions) GetEmbeddings() bool {
-	if x != nil {
-		return x.Embeddings
-	}
-	return false
-}
-
-func (x *ModelOptions) GetNUMA() bool {
-	if x != nil {
-		return x.NUMA
-	}
-	return false
-}
-
-func (x *ModelOptions) GetNGPULayers() int32 {
-	if x != nil {
-		return x.NGPULayers
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetMainGPU() string {
-	if x != nil {
-		return x.MainGPU
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetTensorSplit() string {
-	if x != nil {
-		return x.TensorSplit
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetThreads() int32 {
-	if x != nil {
-		return x.Threads
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetLibrarySearchPath() string {
-	if x != nil {
-		return x.LibrarySearchPath
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetRopeFreqBase() float32 {
-	if x != nil {
-		return x.RopeFreqBase
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetRopeFreqScale() float32 {
-	if x != nil {
-		return x.RopeFreqScale
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetRMSNormEps() float32 {
-	if x != nil {
-		return x.RMSNormEps
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetNGQA() int32 {
-	if x != nil {
-		return x.NGQA
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetModelFile() string {
-	if x != nil {
-		return x.ModelFile
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetDevice() string {
-	if x != nil {
-		return x.Device
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetUseTriton() bool {
-	if x != nil {
-		return x.UseTriton
-	}
-	return false
-}
-
-func (x *ModelOptions) GetModelBaseName() string {
-	if x != nil {
-		return x.ModelBaseName
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetUseFastTokenizer() bool {
-	if x != nil {
-		return x.UseFastTokenizer
-	}
-	return false
-}
-
-func (x *ModelOptions) GetPipelineType() string {
-	if x != nil {
-		return x.PipelineType
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetSchedulerType() string {
-	if x != nil {
-		return x.SchedulerType
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetCUDA() bool {
-	if x != nil {
-		return x.CUDA
-	}
-	return false
-}
-
-func (x *ModelOptions) GetCFGScale() float32 {
-	if x != nil {
-		return x.CFGScale
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetIMG2IMG() bool {
-	if x != nil {
-		return x.IMG2IMG
-	}
-	return false
-}
-
-func (x *ModelOptions) GetCLIPModel() string {
-	if x != nil {
-		return x.CLIPModel
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetCLIPSubfolder() string {
-	if x != nil {
-		return x.CLIPSubfolder
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetCLIPSkip() int32 {
-	if x != nil {
-		return x.CLIPSkip
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetControlNet() string {
-	if x != nil {
-		return x.ControlNet
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetTokenizer() string {
-	if x != nil {
-		return x.Tokenizer
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetLoraBase() string {
-	if x != nil {
-		return x.LoraBase
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetLoraAdapter() string {
-	if x != nil {
-		return x.LoraAdapter
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetLoraScale() float32 {
-	if x != nil {
-		return x.LoraScale
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetNoMulMatQ() bool {
-	if x != nil {
-		return x.NoMulMatQ
-	}
-	return false
-}
-
-func (x *ModelOptions) GetDraftModel() string {
-	if x != nil {
-		return x.DraftModel
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetAudioPath() string {
-	if x != nil {
-		return x.AudioPath
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetQuantization() string {
-	if x != nil {
-		return x.Quantization
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetGPUMemoryUtilization() float32 {
-	if x != nil {
-		return x.GPUMemoryUtilization
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetTrustRemoteCode() bool {
-	if x != nil {
-		return x.TrustRemoteCode
-	}
-	return false
-}
-
-func (x *ModelOptions) GetEnforceEager() bool {
-	if x != nil {
-		return x.EnforceEager
-	}
-	return false
-}
-
-func (x *ModelOptions) GetSwapSpace() int32 {
-	if x != nil {
-		return x.SwapSpace
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetMaxModelLen() int32 {
-	if x != nil {
-		return x.MaxModelLen
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetMMProj() string {
-	if x != nil {
-		return x.MMProj
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetRopeScaling() string {
-	if x != nil {
-		return x.RopeScaling
-	}
-	return ""
-}
-
-func (x *ModelOptions) GetYarnExtFactor() float32 {
-	if x != nil {
-		return x.YarnExtFactor
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetYarnAttnFactor() float32 {
-	if x != nil {
-		return x.YarnAttnFactor
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetYarnBetaFast() float32 {
-	if x != nil {
-		return x.YarnBetaFast
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetYarnBetaSlow() float32 {
-	if x != nil {
-		return x.YarnBetaSlow
-	}
-	return 0
-}
-
-func (x *ModelOptions) GetType() string {
-	if x != nil {
-		return x.Type
-	}
-	return ""
-}
-
-type Result struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Message string `protobuf:"bytes,1,opt,name=message,proto3" json:"message,omitempty"`
-	Success bool   `protobuf:"varint,2,opt,name=success,proto3" json:"success,omitempty"`
-}
-
-func (x *Result) Reset() {
-	*x = Result{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[12]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *Result) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*Result) ProtoMessage() {}
-
-func (x *Result) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[12]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use Result.ProtoReflect.Descriptor instead.
-func (*Result) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{12}
-}
-
-func (x *Result) GetMessage() string {
-	if x != nil {
-		return x.Message
-	}
-	return ""
-}
-
-func (x *Result) GetSuccess() bool {
-	if x != nil {
-		return x.Success
-	}
-	return false
-}
-
-type EmbeddingResult struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Embeddings []float32 `protobuf:"fixed32,1,rep,packed,name=embeddings,proto3" json:"embeddings,omitempty"`
-}
-
-func (x *EmbeddingResult) Reset() {
-	*x = EmbeddingResult{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[13]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *EmbeddingResult) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*EmbeddingResult) ProtoMessage() {}
-
-func (x *EmbeddingResult) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[13]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use EmbeddingResult.ProtoReflect.Descriptor instead.
-func (*EmbeddingResult) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{13}
-}
-
-func (x *EmbeddingResult) GetEmbeddings() []float32 {
-	if x != nil {
-		return x.Embeddings
-	}
-	return nil
-}
-
-type TranscriptRequest struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Dst      string `protobuf:"bytes,2,opt,name=dst,proto3" json:"dst,omitempty"`
-	Language string `protobuf:"bytes,3,opt,name=language,proto3" json:"language,omitempty"`
-	Threads  uint32 `protobuf:"varint,4,opt,name=threads,proto3" json:"threads,omitempty"`
-}
-
-func (x *TranscriptRequest) Reset() {
-	*x = TranscriptRequest{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[14]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *TranscriptRequest) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*TranscriptRequest) ProtoMessage() {}
-
-func (x *TranscriptRequest) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[14]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use TranscriptRequest.ProtoReflect.Descriptor instead.
-func (*TranscriptRequest) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{14}
-}
-
-func (x *TranscriptRequest) GetDst() string {
-	if x != nil {
-		return x.Dst
-	}
-	return ""
-}
-
-func (x *TranscriptRequest) GetLanguage() string {
-	if x != nil {
-		return x.Language
-	}
-	return ""
-}
-
-func (x *TranscriptRequest) GetThreads() uint32 {
-	if x != nil {
-		return x.Threads
-	}
-	return 0
-}
-
-type TranscriptResult struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Segments []*TranscriptSegment `protobuf:"bytes,1,rep,name=segments,proto3" json:"segments,omitempty"`
-	Text     string               `protobuf:"bytes,2,opt,name=text,proto3" json:"text,omitempty"`
-}
-
-func (x *TranscriptResult) Reset() {
-	*x = TranscriptResult{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[15]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *TranscriptResult) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*TranscriptResult) ProtoMessage() {}
-
-func (x *TranscriptResult) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[15]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use TranscriptResult.ProtoReflect.Descriptor instead.
-func (*TranscriptResult) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{15}
-}
-
-func (x *TranscriptResult) GetSegments() []*TranscriptSegment {
-	if x != nil {
-		return x.Segments
-	}
-	return nil
-}
-
-func (x *TranscriptResult) GetText() string {
-	if x != nil {
-		return x.Text
-	}
-	return ""
-}
-
-type TranscriptSegment struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Id     int32   `protobuf:"varint,1,opt,name=id,proto3" json:"id,omitempty"`
-	Start  int64   `protobuf:"varint,2,opt,name=start,proto3" json:"start,omitempty"`
-	End    int64   `protobuf:"varint,3,opt,name=end,proto3" json:"end,omitempty"`
-	Text   string  `protobuf:"bytes,4,opt,name=text,proto3" json:"text,omitempty"`
-	Tokens []int32 `protobuf:"varint,5,rep,packed,name=tokens,proto3" json:"tokens,omitempty"`
-}
-
-func (x *TranscriptSegment) Reset() {
-	*x = TranscriptSegment{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[16]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *TranscriptSegment) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*TranscriptSegment) ProtoMessage() {}
-
-func (x *TranscriptSegment) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[16]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use TranscriptSegment.ProtoReflect.Descriptor instead.
-func (*TranscriptSegment) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{16}
-}
-
-func (x *TranscriptSegment) GetId() int32 {
-	if x != nil {
-		return x.Id
-	}
-	return 0
-}
-
-func (x *TranscriptSegment) GetStart() int64 {
-	if x != nil {
-		return x.Start
-	}
-	return 0
-}
-
-func (x *TranscriptSegment) GetEnd() int64 {
-	if x != nil {
-		return x.End
-	}
-	return 0
-}
-
-func (x *TranscriptSegment) GetText() string {
-	if x != nil {
-		return x.Text
-	}
-	return ""
-}
-
-func (x *TranscriptSegment) GetTokens() []int32 {
-	if x != nil {
-		return x.Tokens
-	}
-	return nil
-}
-
-type GenerateImageRequest struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Height         int32  `protobuf:"varint,1,opt,name=height,proto3" json:"height,omitempty"`
-	Width          int32  `protobuf:"varint,2,opt,name=width,proto3" json:"width,omitempty"`
-	Mode           int32  `protobuf:"varint,3,opt,name=mode,proto3" json:"mode,omitempty"`
-	Step           int32  `protobuf:"varint,4,opt,name=step,proto3" json:"step,omitempty"`
-	Seed           int32  `protobuf:"varint,5,opt,name=seed,proto3" json:"seed,omitempty"`
-	PositivePrompt string `protobuf:"bytes,6,opt,name=positive_prompt,json=positivePrompt,proto3" json:"positive_prompt,omitempty"`
-	NegativePrompt string `protobuf:"bytes,7,opt,name=negative_prompt,json=negativePrompt,proto3" json:"negative_prompt,omitempty"`
-	Dst            string `protobuf:"bytes,8,opt,name=dst,proto3" json:"dst,omitempty"`
-	Src            string `protobuf:"bytes,9,opt,name=src,proto3" json:"src,omitempty"`
-	// Diffusers
-	EnableParameters string `protobuf:"bytes,10,opt,name=EnableParameters,proto3" json:"EnableParameters,omitempty"`
-	CLIPSkip         int32  `protobuf:"varint,11,opt,name=CLIPSkip,proto3" json:"CLIPSkip,omitempty"`
-}
-
-func (x *GenerateImageRequest) Reset() {
-	*x = GenerateImageRequest{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[17]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *GenerateImageRequest) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*GenerateImageRequest) ProtoMessage() {}
-
-func (x *GenerateImageRequest) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[17]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use GenerateImageRequest.ProtoReflect.Descriptor instead.
-func (*GenerateImageRequest) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{17}
-}
-
-func (x *GenerateImageRequest) GetHeight() int32 {
-	if x != nil {
-		return x.Height
-	}
-	return 0
-}
-
-func (x *GenerateImageRequest) GetWidth() int32 {
-	if x != nil {
-		return x.Width
-	}
-	return 0
-}
-
-func (x *GenerateImageRequest) GetMode() int32 {
-	if x != nil {
-		return x.Mode
-	}
-	return 0
-}
-
-func (x *GenerateImageRequest) GetStep() int32 {
-	if x != nil {
-		return x.Step
-	}
-	return 0
-}
-
-func (x *GenerateImageRequest) GetSeed() int32 {
-	if x != nil {
-		return x.Seed
-	}
-	return 0
-}
-
-func (x *GenerateImageRequest) GetPositivePrompt() string {
-	if x != nil {
-		return x.PositivePrompt
-	}
-	return ""
-}
-
-func (x *GenerateImageRequest) GetNegativePrompt() string {
-	if x != nil {
-		return x.NegativePrompt
-	}
-	return ""
-}
-
-func (x *GenerateImageRequest) GetDst() string {
-	if x != nil {
-		return x.Dst
-	}
-	return ""
-}
-
-func (x *GenerateImageRequest) GetSrc() string {
-	if x != nil {
-		return x.Src
-	}
-	return ""
-}
-
-func (x *GenerateImageRequest) GetEnableParameters() string {
-	if x != nil {
-		return x.EnableParameters
-	}
-	return ""
-}
-
-func (x *GenerateImageRequest) GetCLIPSkip() int32 {
-	if x != nil {
-		return x.CLIPSkip
-	}
-	return 0
-}
-
-type TTSRequest struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Text  string `protobuf:"bytes,1,opt,name=text,proto3" json:"text,omitempty"`
-	Model string `protobuf:"bytes,2,opt,name=model,proto3" json:"model,omitempty"`
-	Dst   string `protobuf:"bytes,3,opt,name=dst,proto3" json:"dst,omitempty"`
-	Voice string `protobuf:"bytes,4,opt,name=voice,proto3" json:"voice,omitempty"`
-}
-
-func (x *TTSRequest) Reset() {
-	*x = TTSRequest{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[18]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *TTSRequest) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*TTSRequest) ProtoMessage() {}
-
-func (x *TTSRequest) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[18]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use TTSRequest.ProtoReflect.Descriptor instead.
-func (*TTSRequest) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{18}
-}
-
-func (x *TTSRequest) GetText() string {
-	if x != nil {
-		return x.Text
-	}
-	return ""
-}
-
-func (x *TTSRequest) GetModel() string {
-	if x != nil {
-		return x.Model
-	}
-	return ""
-}
-
-func (x *TTSRequest) GetDst() string {
-	if x != nil {
-		return x.Dst
-	}
-	return ""
-}
-
-func (x *TTSRequest) GetVoice() string {
-	if x != nil {
-		return x.Voice
-	}
-	return ""
-}
-
-type TokenizationResponse struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Length int32   `protobuf:"varint,1,opt,name=length,proto3" json:"length,omitempty"`
-	Tokens []int32 `protobuf:"varint,2,rep,packed,name=tokens,proto3" json:"tokens,omitempty"`
-}
-
-func (x *TokenizationResponse) Reset() {
-	*x = TokenizationResponse{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[19]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *TokenizationResponse) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*TokenizationResponse) ProtoMessage() {}
-
-func (x *TokenizationResponse) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[19]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use TokenizationResponse.ProtoReflect.Descriptor instead.
-func (*TokenizationResponse) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{19}
-}
-
-func (x *TokenizationResponse) GetLength() int32 {
-	if x != nil {
-		return x.Length
-	}
-	return 0
-}
-
-func (x *TokenizationResponse) GetTokens() []int32 {
-	if x != nil {
-		return x.Tokens
-	}
-	return nil
-}
-
-type MemoryUsageData struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Total     uint64            `protobuf:"varint,1,opt,name=total,proto3" json:"total,omitempty"`
-	Breakdown map[string]uint64 `protobuf:"bytes,2,rep,name=breakdown,proto3" json:"breakdown,omitempty" protobuf_key:"bytes,1,opt,name=key,proto3" protobuf_val:"varint,2,opt,name=value,proto3"`
-}
-
-func (x *MemoryUsageData) Reset() {
-	*x = MemoryUsageData{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[20]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *MemoryUsageData) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*MemoryUsageData) ProtoMessage() {}
-
-func (x *MemoryUsageData) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[20]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use MemoryUsageData.ProtoReflect.Descriptor instead.
-func (*MemoryUsageData) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{20}
-}
-
-func (x *MemoryUsageData) GetTotal() uint64 {
-	if x != nil {
-		return x.Total
-	}
-	return 0
-}
-
-func (x *MemoryUsageData) GetBreakdown() map[string]uint64 {
-	if x != nil {
-		return x.Breakdown
-	}
-	return nil
-}
-
-type StatusResponse struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	State  StatusResponse_State `protobuf:"varint,1,opt,name=state,proto3,enum=backend.StatusResponse_State" json:"state,omitempty"`
-	Memory *MemoryUsageData     `protobuf:"bytes,2,opt,name=memory,proto3" json:"memory,omitempty"`
-}
-
-func (x *StatusResponse) Reset() {
-	*x = StatusResponse{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[21]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *StatusResponse) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*StatusResponse) ProtoMessage() {}
-
-func (x *StatusResponse) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[21]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use StatusResponse.ProtoReflect.Descriptor instead.
-func (*StatusResponse) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{21}
-}
-
-func (x *StatusResponse) GetState() StatusResponse_State {
-	if x != nil {
-		return x.State
-	}
-	return StatusResponse_UNINITIALIZED
-}
-
-func (x *StatusResponse) GetMemory() *MemoryUsageData {
-	if x != nil {
-		return x.Memory
-	}
-	return nil
-}
-
-type Message struct {
-	state         protoimpl.MessageState
-	sizeCache     protoimpl.SizeCache
-	unknownFields protoimpl.UnknownFields
-
-	Role    string `protobuf:"bytes,1,opt,name=role,proto3" json:"role,omitempty"`
-	Content string `protobuf:"bytes,2,opt,name=content,proto3" json:"content,omitempty"`
-}
-
-func (x *Message) Reset() {
-	*x = Message{}
-	if protoimpl.UnsafeEnabled {
-		mi := &file_backend_proto_msgTypes[22]
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		ms.StoreMessageInfo(mi)
-	}
-}
-
-func (x *Message) String() string {
-	return protoimpl.X.MessageStringOf(x)
-}
-
-func (*Message) ProtoMessage() {}
-
-func (x *Message) ProtoReflect() protoreflect.Message {
-	mi := &file_backend_proto_msgTypes[22]
-	if protoimpl.UnsafeEnabled && x != nil {
-		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
-		if ms.LoadMessageInfo() == nil {
-			ms.StoreMessageInfo(mi)
-		}
-		return ms
-	}
-	return mi.MessageOf(x)
-}
-
-// Deprecated: Use Message.ProtoReflect.Descriptor instead.
-func (*Message) Descriptor() ([]byte, []int) {
-	return file_backend_proto_rawDescGZIP(), []int{22}
-}
-
-func (x *Message) GetRole() string {
-	if x != nil {
-		return x.Role
-	}
-	return ""
-}
-
-func (x *Message) GetContent() string {
-	if x != nil {
-		return x.Content
-	}
-	return ""
-}
-
-var File_backend_proto protoreflect.FileDescriptor
-
-var file_backend_proto_rawDesc = []byte{
-	0x0a, 0x0d, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x70, 0x72, 0x6f, 0x74, 0x6f, 0x12,
-	0x07, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x22, 0x23, 0x0a, 0x09, 0x53, 0x74, 0x6f, 0x72,
-	0x65, 0x73, 0x4b, 0x65, 0x79, 0x12, 0x16, 0x0a, 0x06, 0x46, 0x6c, 0x6f, 0x61, 0x74, 0x73, 0x18,
-	0x01, 0x20, 0x03, 0x28, 0x02, 0x52, 0x06, 0x46, 0x6c, 0x6f, 0x61, 0x74, 0x73, 0x22, 0x23, 0x0a,
-	0x0b, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x12, 0x14, 0x0a, 0x05,
-	0x42, 0x79, 0x74, 0x65, 0x73, 0x18, 0x01, 0x20, 0x01, 0x28, 0x0c, 0x52, 0x05, 0x42, 0x79, 0x74,
-	0x65, 0x73, 0x22, 0x68, 0x0a, 0x10, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x53, 0x65, 0x74, 0x4f,
-	0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x12, 0x26, 0x0a, 0x04, 0x4b, 0x65, 0x79, 0x73, 0x18, 0x01,
-	0x20, 0x03, 0x28, 0x0b, 0x32, 0x12, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53,
-	0x74, 0x6f, 0x72, 0x65, 0x73, 0x4b, 0x65, 0x79, 0x52, 0x04, 0x4b, 0x65, 0x79, 0x73, 0x12, 0x2c,
-	0x0a, 0x06, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x73, 0x18, 0x02, 0x20, 0x03, 0x28, 0x0b, 0x32, 0x14,
-	0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x56,
-	0x61, 0x6c, 0x75, 0x65, 0x52, 0x06, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x73, 0x22, 0x3d, 0x0a, 0x13,
-	0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x44, 0x65, 0x6c, 0x65, 0x74, 0x65, 0x4f, 0x70, 0x74, 0x69,
-	0x6f, 0x6e, 0x73, 0x12, 0x26, 0x0a, 0x04, 0x4b, 0x65, 0x79, 0x73, 0x18, 0x01, 0x20, 0x03, 0x28,
-	0x0b, 0x32, 0x12, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f, 0x72,
-	0x65, 0x73, 0x4b, 0x65, 0x79, 0x52, 0x04, 0x4b, 0x65, 0x79, 0x73, 0x22, 0x3a, 0x0a, 0x10, 0x53,
-	0x74, 0x6f, 0x72, 0x65, 0x73, 0x47, 0x65, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x12,
-	0x26, 0x0a, 0x04, 0x4b, 0x65, 0x79, 0x73, 0x18, 0x01, 0x20, 0x03, 0x28, 0x0b, 0x32, 0x12, 0x2e,
-	0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x4b, 0x65,
-	0x79, 0x52, 0x04, 0x4b, 0x65, 0x79, 0x73, 0x22, 0x67, 0x0a, 0x0f, 0x53, 0x74, 0x6f, 0x72, 0x65,
-	0x73, 0x47, 0x65, 0x74, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x12, 0x26, 0x0a, 0x04, 0x4b, 0x65,
-	0x79, 0x73, 0x18, 0x01, 0x20, 0x03, 0x28, 0x0b, 0x32, 0x12, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65,
-	0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x4b, 0x65, 0x79, 0x52, 0x04, 0x4b, 0x65,
-	0x79, 0x73, 0x12, 0x2c, 0x0a, 0x06, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x73, 0x18, 0x02, 0x20, 0x03,
-	0x28, 0x0b, 0x32, 0x14, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f,
-	0x72, 0x65, 0x73, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x52, 0x06, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x73,
-	0x22, 0x4d, 0x0a, 0x11, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x46, 0x69, 0x6e, 0x64, 0x4f, 0x70,
-	0x74, 0x69, 0x6f, 0x6e, 0x73, 0x12, 0x24, 0x0a, 0x03, 0x4b, 0x65, 0x79, 0x18, 0x01, 0x20, 0x01,
-	0x28, 0x0b, 0x32, 0x12, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f,
-	0x72, 0x65, 0x73, 0x4b, 0x65, 0x79, 0x52, 0x03, 0x4b, 0x65, 0x79, 0x12, 0x12, 0x0a, 0x04, 0x54,
-	0x6f, 0x70, 0x4b, 0x18, 0x02, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x54, 0x6f, 0x70, 0x4b, 0x22,
-	0x8c, 0x01, 0x0a, 0x10, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x46, 0x69, 0x6e, 0x64, 0x52, 0x65,
-	0x73, 0x75, 0x6c, 0x74, 0x12, 0x26, 0x0a, 0x04, 0x4b, 0x65, 0x79, 0x73, 0x18, 0x01, 0x20, 0x03,
-	0x28, 0x0b, 0x32, 0x12, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f,
-	0x72, 0x65, 0x73, 0x4b, 0x65, 0x79, 0x52, 0x04, 0x4b, 0x65, 0x79, 0x73, 0x12, 0x2c, 0x0a, 0x06,
-	0x56, 0x61, 0x6c, 0x75, 0x65, 0x73, 0x18, 0x02, 0x20, 0x03, 0x28, 0x0b, 0x32, 0x14, 0x2e, 0x62,
-	0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x56, 0x61, 0x6c,
-	0x75, 0x65, 0x52, 0x06, 0x56, 0x61, 0x6c, 0x75, 0x65, 0x73, 0x12, 0x22, 0x0a, 0x0c, 0x53, 0x69,
-	0x6d, 0x69, 0x6c, 0x61, 0x72, 0x69, 0x74, 0x69, 0x65, 0x73, 0x18, 0x03, 0x20, 0x03, 0x28, 0x02,
-	0x52, 0x0c, 0x53, 0x69, 0x6d, 0x69, 0x6c, 0x61, 0x72, 0x69, 0x74, 0x69, 0x65, 0x73, 0x22, 0x0f,
-	0x0a, 0x0d, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x22,
-	0xd6, 0x0a, 0x0a, 0x0e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f,
-	0x6e, 0x73, 0x12, 0x16, 0x0a, 0x06, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x18, 0x01, 0x20, 0x01,
-	0x28, 0x09, 0x52, 0x06, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x12, 0x12, 0x0a, 0x04, 0x53, 0x65,
-	0x65, 0x64, 0x18, 0x02, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x53, 0x65, 0x65, 0x64, 0x12, 0x18,
-	0x0a, 0x07, 0x54, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x18, 0x03, 0x20, 0x01, 0x28, 0x05, 0x52,
-	0x07, 0x54, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x12, 0x16, 0x0a, 0x06, 0x54, 0x6f, 0x6b, 0x65,
-	0x6e, 0x73, 0x18, 0x04, 0x20, 0x01, 0x28, 0x05, 0x52, 0x06, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x73,
-	0x12, 0x12, 0x0a, 0x04, 0x54, 0x6f, 0x70, 0x4b, 0x18, 0x05, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04,
-	0x54, 0x6f, 0x70, 0x4b, 0x12, 0x16, 0x0a, 0x06, 0x52, 0x65, 0x70, 0x65, 0x61, 0x74, 0x18, 0x06,
-	0x20, 0x01, 0x28, 0x05, 0x52, 0x06, 0x52, 0x65, 0x70, 0x65, 0x61, 0x74, 0x12, 0x14, 0x0a, 0x05,
-	0x42, 0x61, 0x74, 0x63, 0x68, 0x18, 0x07, 0x20, 0x01, 0x28, 0x05, 0x52, 0x05, 0x42, 0x61, 0x74,
-	0x63, 0x68, 0x12, 0x14, 0x0a, 0x05, 0x4e, 0x4b, 0x65, 0x65, 0x70, 0x18, 0x08, 0x20, 0x01, 0x28,
-	0x05, 0x52, 0x05, 0x4e, 0x4b, 0x65, 0x65, 0x70, 0x12, 0x20, 0x0a, 0x0b, 0x54, 0x65, 0x6d, 0x70,
-	0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x18, 0x09, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0b, 0x54,
-	0x65, 0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x12, 0x18, 0x0a, 0x07, 0x50, 0x65,
-	0x6e, 0x61, 0x6c, 0x74, 0x79, 0x18, 0x0a, 0x20, 0x01, 0x28, 0x02, 0x52, 0x07, 0x50, 0x65, 0x6e,
-	0x61, 0x6c, 0x74, 0x79, 0x12, 0x14, 0x0a, 0x05, 0x46, 0x31, 0x36, 0x4b, 0x56, 0x18, 0x0b, 0x20,
-	0x01, 0x28, 0x08, 0x52, 0x05, 0x46, 0x31, 0x36, 0x4b, 0x56, 0x12, 0x1c, 0x0a, 0x09, 0x44, 0x65,
-	0x62, 0x75, 0x67, 0x4d, 0x6f, 0x64, 0x65, 0x18, 0x0c, 0x20, 0x01, 0x28, 0x08, 0x52, 0x09, 0x44,
-	0x65, 0x62, 0x75, 0x67, 0x4d, 0x6f, 0x64, 0x65, 0x12, 0x20, 0x0a, 0x0b, 0x53, 0x74, 0x6f, 0x70,
-	0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x73, 0x18, 0x0d, 0x20, 0x03, 0x28, 0x09, 0x52, 0x0b, 0x53,
-	0x74, 0x6f, 0x70, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x73, 0x12, 0x1c, 0x0a, 0x09, 0x49, 0x67,
-	0x6e, 0x6f, 0x72, 0x65, 0x45, 0x4f, 0x53, 0x18, 0x0e, 0x20, 0x01, 0x28, 0x08, 0x52, 0x09, 0x49,
-	0x67, 0x6e, 0x6f, 0x72, 0x65, 0x45, 0x4f, 0x53, 0x12, 0x2c, 0x0a, 0x11, 0x54, 0x61, 0x69, 0x6c,
-	0x46, 0x72, 0x65, 0x65, 0x53, 0x61, 0x6d, 0x70, 0x6c, 0x69, 0x6e, 0x67, 0x5a, 0x18, 0x0f, 0x20,
-	0x01, 0x28, 0x02, 0x52, 0x11, 0x54, 0x61, 0x69, 0x6c, 0x46, 0x72, 0x65, 0x65, 0x53, 0x61, 0x6d,
-	0x70, 0x6c, 0x69, 0x6e, 0x67, 0x5a, 0x12, 0x1a, 0x0a, 0x08, 0x54, 0x79, 0x70, 0x69, 0x63, 0x61,
-	0x6c, 0x50, 0x18, 0x10, 0x20, 0x01, 0x28, 0x02, 0x52, 0x08, 0x54, 0x79, 0x70, 0x69, 0x63, 0x61,
-	0x6c, 0x50, 0x12, 0x2a, 0x0a, 0x10, 0x46, 0x72, 0x65, 0x71, 0x75, 0x65, 0x6e, 0x63, 0x79, 0x50,
-	0x65, 0x6e, 0x61, 0x6c, 0x74, 0x79, 0x18, 0x11, 0x20, 0x01, 0x28, 0x02, 0x52, 0x10, 0x46, 0x72,
-	0x65, 0x71, 0x75, 0x65, 0x6e, 0x63, 0x79, 0x50, 0x65, 0x6e, 0x61, 0x6c, 0x74, 0x79, 0x12, 0x28,
-	0x0a, 0x0f, 0x50, 0x72, 0x65, 0x73, 0x65, 0x6e, 0x63, 0x65, 0x50, 0x65, 0x6e, 0x61, 0x6c, 0x74,
-	0x79, 0x18, 0x12, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0f, 0x50, 0x72, 0x65, 0x73, 0x65, 0x6e, 0x63,
-	0x65, 0x50, 0x65, 0x6e, 0x61, 0x6c, 0x74, 0x79, 0x12, 0x1a, 0x0a, 0x08, 0x4d, 0x69, 0x72, 0x6f,
-	0x73, 0x74, 0x61, 0x74, 0x18, 0x13, 0x20, 0x01, 0x28, 0x05, 0x52, 0x08, 0x4d, 0x69, 0x72, 0x6f,
-	0x73, 0x74, 0x61, 0x74, 0x12, 0x20, 0x0a, 0x0b, 0x4d, 0x69, 0x72, 0x6f, 0x73, 0x74, 0x61, 0x74,
-	0x45, 0x54, 0x41, 0x18, 0x14, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0b, 0x4d, 0x69, 0x72, 0x6f, 0x73,
-	0x74, 0x61, 0x74, 0x45, 0x54, 0x41, 0x12, 0x20, 0x0a, 0x0b, 0x4d, 0x69, 0x72, 0x6f, 0x73, 0x74,
-	0x61, 0x74, 0x54, 0x41, 0x55, 0x18, 0x15, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0b, 0x4d, 0x69, 0x72,
-	0x6f, 0x73, 0x74, 0x61, 0x74, 0x54, 0x41, 0x55, 0x12, 0x1e, 0x0a, 0x0a, 0x50, 0x65, 0x6e, 0x61,
-	0x6c, 0x69, 0x7a, 0x65, 0x4e, 0x4c, 0x18, 0x16, 0x20, 0x01, 0x28, 0x08, 0x52, 0x0a, 0x50, 0x65,
-	0x6e, 0x61, 0x6c, 0x69, 0x7a, 0x65, 0x4e, 0x4c, 0x12, 0x1c, 0x0a, 0x09, 0x4c, 0x6f, 0x67, 0x69,
-	0x74, 0x42, 0x69, 0x61, 0x73, 0x18, 0x17, 0x20, 0x01, 0x28, 0x09, 0x52, 0x09, 0x4c, 0x6f, 0x67,
-	0x69, 0x74, 0x42, 0x69, 0x61, 0x73, 0x12, 0x14, 0x0a, 0x05, 0x4d, 0x4c, 0x6f, 0x63, 0x6b, 0x18,
-	0x19, 0x20, 0x01, 0x28, 0x08, 0x52, 0x05, 0x4d, 0x4c, 0x6f, 0x63, 0x6b, 0x12, 0x12, 0x0a, 0x04,
-	0x4d, 0x4d, 0x61, 0x70, 0x18, 0x1a, 0x20, 0x01, 0x28, 0x08, 0x52, 0x04, 0x4d, 0x4d, 0x61, 0x70,
-	0x12, 0x26, 0x0a, 0x0e, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x43, 0x61, 0x63, 0x68, 0x65, 0x41,
-	0x6c, 0x6c, 0x18, 0x1b, 0x20, 0x01, 0x28, 0x08, 0x52, 0x0e, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74,
-	0x43, 0x61, 0x63, 0x68, 0x65, 0x41, 0x6c, 0x6c, 0x12, 0x24, 0x0a, 0x0d, 0x50, 0x72, 0x6f, 0x6d,
-	0x70, 0x74, 0x43, 0x61, 0x63, 0x68, 0x65, 0x52, 0x4f, 0x18, 0x1c, 0x20, 0x01, 0x28, 0x08, 0x52,
-	0x0d, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x43, 0x61, 0x63, 0x68, 0x65, 0x52, 0x4f, 0x12, 0x18,
-	0x0a, 0x07, 0x47, 0x72, 0x61, 0x6d, 0x6d, 0x61, 0x72, 0x18, 0x1d, 0x20, 0x01, 0x28, 0x09, 0x52,
-	0x07, 0x47, 0x72, 0x61, 0x6d, 0x6d, 0x61, 0x72, 0x12, 0x18, 0x0a, 0x07, 0x4d, 0x61, 0x69, 0x6e,
-	0x47, 0x50, 0x55, 0x18, 0x1e, 0x20, 0x01, 0x28, 0x09, 0x52, 0x07, 0x4d, 0x61, 0x69, 0x6e, 0x47,
-	0x50, 0x55, 0x12, 0x20, 0x0a, 0x0b, 0x54, 0x65, 0x6e, 0x73, 0x6f, 0x72, 0x53, 0x70, 0x6c, 0x69,
-	0x74, 0x18, 0x1f, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0b, 0x54, 0x65, 0x6e, 0x73, 0x6f, 0x72, 0x53,
-	0x70, 0x6c, 0x69, 0x74, 0x12, 0x12, 0x0a, 0x04, 0x54, 0x6f, 0x70, 0x50, 0x18, 0x20, 0x20, 0x01,
-	0x28, 0x02, 0x52, 0x04, 0x54, 0x6f, 0x70, 0x50, 0x12, 0x28, 0x0a, 0x0f, 0x50, 0x72, 0x6f, 0x6d,
-	0x70, 0x74, 0x43, 0x61, 0x63, 0x68, 0x65, 0x50, 0x61, 0x74, 0x68, 0x18, 0x21, 0x20, 0x01, 0x28,
-	0x09, 0x52, 0x0f, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x43, 0x61, 0x63, 0x68, 0x65, 0x50, 0x61,
-	0x74, 0x68, 0x12, 0x14, 0x0a, 0x05, 0x44, 0x65, 0x62, 0x75, 0x67, 0x18, 0x22, 0x20, 0x01, 0x28,
-	0x08, 0x52, 0x05, 0x44, 0x65, 0x62, 0x75, 0x67, 0x12, 0x28, 0x0a, 0x0f, 0x45, 0x6d, 0x62, 0x65,
-	0x64, 0x64, 0x69, 0x6e, 0x67, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x73, 0x18, 0x23, 0x20, 0x03, 0x28,
-	0x05, 0x52, 0x0f, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x54, 0x6f, 0x6b, 0x65,
-	0x6e, 0x73, 0x12, 0x1e, 0x0a, 0x0a, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73,
-	0x18, 0x24, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0a, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e,
-	0x67, 0x73, 0x12, 0x22, 0x0a, 0x0c, 0x52, 0x6f, 0x70, 0x65, 0x46, 0x72, 0x65, 0x71, 0x42, 0x61,
-	0x73, 0x65, 0x18, 0x25, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0c, 0x52, 0x6f, 0x70, 0x65, 0x46, 0x72,
-	0x65, 0x71, 0x42, 0x61, 0x73, 0x65, 0x12, 0x24, 0x0a, 0x0d, 0x52, 0x6f, 0x70, 0x65, 0x46, 0x72,
-	0x65, 0x71, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x18, 0x26, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0d, 0x52,
-	0x6f, 0x70, 0x65, 0x46, 0x72, 0x65, 0x71, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x12, 0x30, 0x0a, 0x13,
-	0x4e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x53, 0x63,
-	0x61, 0x6c, 0x65, 0x18, 0x27, 0x20, 0x01, 0x28, 0x02, 0x52, 0x13, 0x4e, 0x65, 0x67, 0x61, 0x74,
-	0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x12, 0x26,
-	0x0a, 0x0e, 0x4e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74,
-	0x18, 0x28, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0e, 0x4e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65,
-	0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x12, 0x16, 0x0a, 0x06, 0x4e, 0x44, 0x72, 0x61, 0x66, 0x74,
-	0x18, 0x29, 0x20, 0x01, 0x28, 0x05, 0x52, 0x06, 0x4e, 0x44, 0x72, 0x61, 0x66, 0x74, 0x12, 0x16,
-	0x0a, 0x06, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x73, 0x18, 0x2a, 0x20, 0x03, 0x28, 0x09, 0x52, 0x06,
-	0x49, 0x6d, 0x61, 0x67, 0x65, 0x73, 0x12, 0x32, 0x0a, 0x14, 0x55, 0x73, 0x65, 0x54, 0x6f, 0x6b,
-	0x65, 0x6e, 0x69, 0x7a, 0x65, 0x72, 0x54, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x18, 0x2b,
-	0x20, 0x01, 0x28, 0x08, 0x52, 0x14, 0x55, 0x73, 0x65, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x69, 0x7a,
-	0x65, 0x72, 0x54, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x12, 0x2c, 0x0a, 0x08, 0x4d, 0x65,
-	0x73, 0x73, 0x61, 0x67, 0x65, 0x73, 0x18, 0x2c, 0x20, 0x03, 0x28, 0x0b, 0x32, 0x10, 0x2e, 0x62,
-	0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x52, 0x08,
-	0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x73, 0x22, 0x21, 0x0a, 0x05, 0x52, 0x65, 0x70, 0x6c,
-	0x79, 0x12, 0x18, 0x0a, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x18, 0x01, 0x20, 0x01,
-	0x28, 0x0c, 0x52, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x22, 0xa2, 0x0d, 0x0a, 0x0c,
-	0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x12, 0x14, 0x0a, 0x05,
-	0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x05, 0x4d, 0x6f, 0x64,
-	0x65, 0x6c, 0x12, 0x20, 0x0a, 0x0b, 0x43, 0x6f, 0x6e, 0x74, 0x65, 0x78, 0x74, 0x53, 0x69, 0x7a,
-	0x65, 0x18, 0x02, 0x20, 0x01, 0x28, 0x05, 0x52, 0x0b, 0x43, 0x6f, 0x6e, 0x74, 0x65, 0x78, 0x74,
-	0x53, 0x69, 0x7a, 0x65, 0x12, 0x12, 0x0a, 0x04, 0x53, 0x65, 0x65, 0x64, 0x18, 0x03, 0x20, 0x01,
-	0x28, 0x05, 0x52, 0x04, 0x53, 0x65, 0x65, 0x64, 0x12, 0x16, 0x0a, 0x06, 0x4e, 0x42, 0x61, 0x74,
-	0x63, 0x68, 0x18, 0x04, 0x20, 0x01, 0x28, 0x05, 0x52, 0x06, 0x4e, 0x42, 0x61, 0x74, 0x63, 0x68,
-	0x12, 0x1c, 0x0a, 0x09, 0x46, 0x31, 0x36, 0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x18, 0x05, 0x20,
-	0x01, 0x28, 0x08, 0x52, 0x09, 0x46, 0x31, 0x36, 0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x12, 0x14,
-	0x0a, 0x05, 0x4d, 0x4c, 0x6f, 0x63, 0x6b, 0x18, 0x06, 0x20, 0x01, 0x28, 0x08, 0x52, 0x05, 0x4d,
-	0x4c, 0x6f, 0x63, 0x6b, 0x12, 0x12, 0x0a, 0x04, 0x4d, 0x4d, 0x61, 0x70, 0x18, 0x07, 0x20, 0x01,
-	0x28, 0x08, 0x52, 0x04, 0x4d, 0x4d, 0x61, 0x70, 0x12, 0x1c, 0x0a, 0x09, 0x56, 0x6f, 0x63, 0x61,
-	0x62, 0x4f, 0x6e, 0x6c, 0x79, 0x18, 0x08, 0x20, 0x01, 0x28, 0x08, 0x52, 0x09, 0x56, 0x6f, 0x63,
-	0x61, 0x62, 0x4f, 0x6e, 0x6c, 0x79, 0x12, 0x18, 0x0a, 0x07, 0x4c, 0x6f, 0x77, 0x56, 0x52, 0x41,
-	0x4d, 0x18, 0x09, 0x20, 0x01, 0x28, 0x08, 0x52, 0x07, 0x4c, 0x6f, 0x77, 0x56, 0x52, 0x41, 0x4d,
-	0x12, 0x1e, 0x0a, 0x0a, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73, 0x18, 0x0a,
-	0x20, 0x01, 0x28, 0x08, 0x52, 0x0a, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73,
-	0x12, 0x12, 0x0a, 0x04, 0x4e, 0x55, 0x4d, 0x41, 0x18, 0x0b, 0x20, 0x01, 0x28, 0x08, 0x52, 0x04,
-	0x4e, 0x55, 0x4d, 0x41, 0x12, 0x1e, 0x0a, 0x0a, 0x4e, 0x47, 0x50, 0x55, 0x4c, 0x61, 0x79, 0x65,
-	0x72, 0x73, 0x18, 0x0c, 0x20, 0x01, 0x28, 0x05, 0x52, 0x0a, 0x4e, 0x47, 0x50, 0x55, 0x4c, 0x61,
-	0x79, 0x65, 0x72, 0x73, 0x12, 0x18, 0x0a, 0x07, 0x4d, 0x61, 0x69, 0x6e, 0x47, 0x50, 0x55, 0x18,
-	0x0d, 0x20, 0x01, 0x28, 0x09, 0x52, 0x07, 0x4d, 0x61, 0x69, 0x6e, 0x47, 0x50, 0x55, 0x12, 0x20,
-	0x0a, 0x0b, 0x54, 0x65, 0x6e, 0x73, 0x6f, 0x72, 0x53, 0x70, 0x6c, 0x69, 0x74, 0x18, 0x0e, 0x20,
-	0x01, 0x28, 0x09, 0x52, 0x0b, 0x54, 0x65, 0x6e, 0x73, 0x6f, 0x72, 0x53, 0x70, 0x6c, 0x69, 0x74,
-	0x12, 0x18, 0x0a, 0x07, 0x54, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x18, 0x0f, 0x20, 0x01, 0x28,
-	0x05, 0x52, 0x07, 0x54, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x12, 0x2c, 0x0a, 0x11, 0x4c, 0x69,
-	0x62, 0x72, 0x61, 0x72, 0x79, 0x53, 0x65, 0x61, 0x72, 0x63, 0x68, 0x50, 0x61, 0x74, 0x68, 0x18,
-	0x10, 0x20, 0x01, 0x28, 0x09, 0x52, 0x11, 0x4c, 0x69, 0x62, 0x72, 0x61, 0x72, 0x79, 0x53, 0x65,
-	0x61, 0x72, 0x63, 0x68, 0x50, 0x61, 0x74, 0x68, 0x12, 0x22, 0x0a, 0x0c, 0x52, 0x6f, 0x70, 0x65,
-	0x46, 0x72, 0x65, 0x71, 0x42, 0x61, 0x73, 0x65, 0x18, 0x11, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0c,
-	0x52, 0x6f, 0x70, 0x65, 0x46, 0x72, 0x65, 0x71, 0x42, 0x61, 0x73, 0x65, 0x12, 0x24, 0x0a, 0x0d,
-	0x52, 0x6f, 0x70, 0x65, 0x46, 0x72, 0x65, 0x71, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x18, 0x12, 0x20,
-	0x01, 0x28, 0x02, 0x52, 0x0d, 0x52, 0x6f, 0x70, 0x65, 0x46, 0x72, 0x65, 0x71, 0x53, 0x63, 0x61,
-	0x6c, 0x65, 0x12, 0x1e, 0x0a, 0x0a, 0x52, 0x4d, 0x53, 0x4e, 0x6f, 0x72, 0x6d, 0x45, 0x70, 0x73,
-	0x18, 0x13, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0a, 0x52, 0x4d, 0x53, 0x4e, 0x6f, 0x72, 0x6d, 0x45,
-	0x70, 0x73, 0x12, 0x12, 0x0a, 0x04, 0x4e, 0x47, 0x51, 0x41, 0x18, 0x14, 0x20, 0x01, 0x28, 0x05,
-	0x52, 0x04, 0x4e, 0x47, 0x51, 0x41, 0x12, 0x1c, 0x0a, 0x09, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x46,
-	0x69, 0x6c, 0x65, 0x18, 0x15, 0x20, 0x01, 0x28, 0x09, 0x52, 0x09, 0x4d, 0x6f, 0x64, 0x65, 0x6c,
-	0x46, 0x69, 0x6c, 0x65, 0x12, 0x16, 0x0a, 0x06, 0x44, 0x65, 0x76, 0x69, 0x63, 0x65, 0x18, 0x16,
-	0x20, 0x01, 0x28, 0x09, 0x52, 0x06, 0x44, 0x65, 0x76, 0x69, 0x63, 0x65, 0x12, 0x1c, 0x0a, 0x09,
-	0x55, 0x73, 0x65, 0x54, 0x72, 0x69, 0x74, 0x6f, 0x6e, 0x18, 0x17, 0x20, 0x01, 0x28, 0x08, 0x52,
-	0x09, 0x55, 0x73, 0x65, 0x54, 0x72, 0x69, 0x74, 0x6f, 0x6e, 0x12, 0x24, 0x0a, 0x0d, 0x4d, 0x6f,
-	0x64, 0x65, 0x6c, 0x42, 0x61, 0x73, 0x65, 0x4e, 0x61, 0x6d, 0x65, 0x18, 0x18, 0x20, 0x01, 0x28,
-	0x09, 0x52, 0x0d, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x42, 0x61, 0x73, 0x65, 0x4e, 0x61, 0x6d, 0x65,
-	0x12, 0x2a, 0x0a, 0x10, 0x55, 0x73, 0x65, 0x46, 0x61, 0x73, 0x74, 0x54, 0x6f, 0x6b, 0x65, 0x6e,
-	0x69, 0x7a, 0x65, 0x72, 0x18, 0x19, 0x20, 0x01, 0x28, 0x08, 0x52, 0x10, 0x55, 0x73, 0x65, 0x46,
-	0x61, 0x73, 0x74, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x69, 0x7a, 0x65, 0x72, 0x12, 0x22, 0x0a, 0x0c,
-	0x50, 0x69, 0x70, 0x65, 0x6c, 0x69, 0x6e, 0x65, 0x54, 0x79, 0x70, 0x65, 0x18, 0x1a, 0x20, 0x01,
-	0x28, 0x09, 0x52, 0x0c, 0x50, 0x69, 0x70, 0x65, 0x6c, 0x69, 0x6e, 0x65, 0x54, 0x79, 0x70, 0x65,
-	0x12, 0x24, 0x0a, 0x0d, 0x53, 0x63, 0x68, 0x65, 0x64, 0x75, 0x6c, 0x65, 0x72, 0x54, 0x79, 0x70,
-	0x65, 0x18, 0x1b, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0d, 0x53, 0x63, 0x68, 0x65, 0x64, 0x75, 0x6c,
-	0x65, 0x72, 0x54, 0x79, 0x70, 0x65, 0x12, 0x12, 0x0a, 0x04, 0x43, 0x55, 0x44, 0x41, 0x18, 0x1c,
-	0x20, 0x01, 0x28, 0x08, 0x52, 0x04, 0x43, 0x55, 0x44, 0x41, 0x12, 0x1a, 0x0a, 0x08, 0x43, 0x46,
-	0x47, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x18, 0x1d, 0x20, 0x01, 0x28, 0x02, 0x52, 0x08, 0x43, 0x46,
-	0x47, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x12, 0x18, 0x0a, 0x07, 0x49, 0x4d, 0x47, 0x32, 0x49, 0x4d,
-	0x47, 0x18, 0x1e, 0x20, 0x01, 0x28, 0x08, 0x52, 0x07, 0x49, 0x4d, 0x47, 0x32, 0x49, 0x4d, 0x47,
-	0x12, 0x1c, 0x0a, 0x09, 0x43, 0x4c, 0x49, 0x50, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x18, 0x1f, 0x20,
-	0x01, 0x28, 0x09, 0x52, 0x09, 0x43, 0x4c, 0x49, 0x50, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x24,
-	0x0a, 0x0d, 0x43, 0x4c, 0x49, 0x50, 0x53, 0x75, 0x62, 0x66, 0x6f, 0x6c, 0x64, 0x65, 0x72, 0x18,
-	0x20, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0d, 0x43, 0x4c, 0x49, 0x50, 0x53, 0x75, 0x62, 0x66, 0x6f,
-	0x6c, 0x64, 0x65, 0x72, 0x12, 0x1a, 0x0a, 0x08, 0x43, 0x4c, 0x49, 0x50, 0x53, 0x6b, 0x69, 0x70,
-	0x18, 0x21, 0x20, 0x01, 0x28, 0x05, 0x52, 0x08, 0x43, 0x4c, 0x49, 0x50, 0x53, 0x6b, 0x69, 0x70,
-	0x12, 0x1e, 0x0a, 0x0a, 0x43, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x4e, 0x65, 0x74, 0x18, 0x30,
-	0x20, 0x01, 0x28, 0x09, 0x52, 0x0a, 0x43, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x4e, 0x65, 0x74,
-	0x12, 0x1c, 0x0a, 0x09, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x69, 0x7a, 0x65, 0x72, 0x18, 0x22, 0x20,
-	0x01, 0x28, 0x09, 0x52, 0x09, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x69, 0x7a, 0x65, 0x72, 0x12, 0x1a,
-	0x0a, 0x08, 0x4c, 0x6f, 0x72, 0x61, 0x42, 0x61, 0x73, 0x65, 0x18, 0x23, 0x20, 0x01, 0x28, 0x09,
-	0x52, 0x08, 0x4c, 0x6f, 0x72, 0x61, 0x42, 0x61, 0x73, 0x65, 0x12, 0x20, 0x0a, 0x0b, 0x4c, 0x6f,
-	0x72, 0x61, 0x41, 0x64, 0x61, 0x70, 0x74, 0x65, 0x72, 0x18, 0x24, 0x20, 0x01, 0x28, 0x09, 0x52,
-	0x0b, 0x4c, 0x6f, 0x72, 0x61, 0x41, 0x64, 0x61, 0x70, 0x74, 0x65, 0x72, 0x12, 0x1c, 0x0a, 0x09,
-	0x4c, 0x6f, 0x72, 0x61, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x18, 0x2a, 0x20, 0x01, 0x28, 0x02, 0x52,
-	0x09, 0x4c, 0x6f, 0x72, 0x61, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x12, 0x1c, 0x0a, 0x09, 0x4e, 0x6f,
-	0x4d, 0x75, 0x6c, 0x4d, 0x61, 0x74, 0x51, 0x18, 0x25, 0x20, 0x01, 0x28, 0x08, 0x52, 0x09, 0x4e,
-	0x6f, 0x4d, 0x75, 0x6c, 0x4d, 0x61, 0x74, 0x51, 0x12, 0x1e, 0x0a, 0x0a, 0x44, 0x72, 0x61, 0x66,
-	0x74, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x18, 0x27, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0a, 0x44, 0x72,
-	0x61, 0x66, 0x74, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x1c, 0x0a, 0x09, 0x41, 0x75, 0x64, 0x69,
-	0x6f, 0x50, 0x61, 0x74, 0x68, 0x18, 0x26, 0x20, 0x01, 0x28, 0x09, 0x52, 0x09, 0x41, 0x75, 0x64,
-	0x69, 0x6f, 0x50, 0x61, 0x74, 0x68, 0x12, 0x22, 0x0a, 0x0c, 0x51, 0x75, 0x61, 0x6e, 0x74, 0x69,
-	0x7a, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x18, 0x28, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0c, 0x51, 0x75,
-	0x61, 0x6e, 0x74, 0x69, 0x7a, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x12, 0x32, 0x0a, 0x14, 0x47, 0x50,
-	0x55, 0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x55, 0x74, 0x69, 0x6c, 0x69, 0x7a, 0x61, 0x74, 0x69,
-	0x6f, 0x6e, 0x18, 0x32, 0x20, 0x01, 0x28, 0x02, 0x52, 0x14, 0x47, 0x50, 0x55, 0x4d, 0x65, 0x6d,
-	0x6f, 0x72, 0x79, 0x55, 0x74, 0x69, 0x6c, 0x69, 0x7a, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x12, 0x28,
-	0x0a, 0x0f, 0x54, 0x72, 0x75, 0x73, 0x74, 0x52, 0x65, 0x6d, 0x6f, 0x74, 0x65, 0x43, 0x6f, 0x64,
-	0x65, 0x18, 0x33, 0x20, 0x01, 0x28, 0x08, 0x52, 0x0f, 0x54, 0x72, 0x75, 0x73, 0x74, 0x52, 0x65,
-	0x6d, 0x6f, 0x74, 0x65, 0x43, 0x6f, 0x64, 0x65, 0x12, 0x22, 0x0a, 0x0c, 0x45, 0x6e, 0x66, 0x6f,
-	0x72, 0x63, 0x65, 0x45, 0x61, 0x67, 0x65, 0x72, 0x18, 0x34, 0x20, 0x01, 0x28, 0x08, 0x52, 0x0c,
-	0x45, 0x6e, 0x66, 0x6f, 0x72, 0x63, 0x65, 0x45, 0x61, 0x67, 0x65, 0x72, 0x12, 0x1c, 0x0a, 0x09,
-	0x53, 0x77, 0x61, 0x70, 0x53, 0x70, 0x61, 0x63, 0x65, 0x18, 0x35, 0x20, 0x01, 0x28, 0x05, 0x52,
-	0x09, 0x53, 0x77, 0x61, 0x70, 0x53, 0x70, 0x61, 0x63, 0x65, 0x12, 0x20, 0x0a, 0x0b, 0x4d, 0x61,
-	0x78, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4c, 0x65, 0x6e, 0x18, 0x36, 0x20, 0x01, 0x28, 0x05, 0x52,
-	0x0b, 0x4d, 0x61, 0x78, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4c, 0x65, 0x6e, 0x12, 0x16, 0x0a, 0x06,
-	0x4d, 0x4d, 0x50, 0x72, 0x6f, 0x6a, 0x18, 0x29, 0x20, 0x01, 0x28, 0x09, 0x52, 0x06, 0x4d, 0x4d,
-	0x50, 0x72, 0x6f, 0x6a, 0x12, 0x20, 0x0a, 0x0b, 0x52, 0x6f, 0x70, 0x65, 0x53, 0x63, 0x61, 0x6c,
-	0x69, 0x6e, 0x67, 0x18, 0x2b, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0b, 0x52, 0x6f, 0x70, 0x65, 0x53,
-	0x63, 0x61, 0x6c, 0x69, 0x6e, 0x67, 0x12, 0x24, 0x0a, 0x0d, 0x59, 0x61, 0x72, 0x6e, 0x45, 0x78,
-	0x74, 0x46, 0x61, 0x63, 0x74, 0x6f, 0x72, 0x18, 0x2c, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0d, 0x59,
-	0x61, 0x72, 0x6e, 0x45, 0x78, 0x74, 0x46, 0x61, 0x63, 0x74, 0x6f, 0x72, 0x12, 0x26, 0x0a, 0x0e,
-	0x59, 0x61, 0x72, 0x6e, 0x41, 0x74, 0x74, 0x6e, 0x46, 0x61, 0x63, 0x74, 0x6f, 0x72, 0x18, 0x2d,
-	0x20, 0x01, 0x28, 0x02, 0x52, 0x0e, 0x59, 0x61, 0x72, 0x6e, 0x41, 0x74, 0x74, 0x6e, 0x46, 0x61,
-	0x63, 0x74, 0x6f, 0x72, 0x12, 0x22, 0x0a, 0x0c, 0x59, 0x61, 0x72, 0x6e, 0x42, 0x65, 0x74, 0x61,
-	0x46, 0x61, 0x73, 0x74, 0x18, 0x2e, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0c, 0x59, 0x61, 0x72, 0x6e,
-	0x42, 0x65, 0x74, 0x61, 0x46, 0x61, 0x73, 0x74, 0x12, 0x22, 0x0a, 0x0c, 0x59, 0x61, 0x72, 0x6e,
-	0x42, 0x65, 0x74, 0x61, 0x53, 0x6c, 0x6f, 0x77, 0x18, 0x2f, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0c,
-	0x59, 0x61, 0x72, 0x6e, 0x42, 0x65, 0x74, 0x61, 0x53, 0x6c, 0x6f, 0x77, 0x12, 0x12, 0x0a, 0x04,
-	0x54, 0x79, 0x70, 0x65, 0x18, 0x31, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x54, 0x79, 0x70, 0x65,
-	0x22, 0x3c, 0x0a, 0x06, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x12, 0x18, 0x0a, 0x07, 0x6d, 0x65,
-	0x73, 0x73, 0x61, 0x67, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x07, 0x6d, 0x65, 0x73,
-	0x73, 0x61, 0x67, 0x65, 0x12, 0x18, 0x0a, 0x07, 0x73, 0x75, 0x63, 0x63, 0x65, 0x73, 0x73, 0x18,
-	0x02, 0x20, 0x01, 0x28, 0x08, 0x52, 0x07, 0x73, 0x75, 0x63, 0x63, 0x65, 0x73, 0x73, 0x22, 0x31,
-	0x0a, 0x0f, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, 0x6c,
-	0x74, 0x12, 0x1e, 0x0a, 0x0a, 0x65, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73, 0x18,
-	0x01, 0x20, 0x03, 0x28, 0x02, 0x52, 0x0a, 0x65, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67,
-	0x73, 0x22, 0x5b, 0x0a, 0x11, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52,
-	0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x02, 0x20,
-	0x01, 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, 0x74, 0x12, 0x1a, 0x0a, 0x08, 0x6c, 0x61, 0x6e, 0x67,
-	0x75, 0x61, 0x67, 0x65, 0x18, 0x03, 0x20, 0x01, 0x28, 0x09, 0x52, 0x08, 0x6c, 0x61, 0x6e, 0x67,
-	0x75, 0x61, 0x67, 0x65, 0x12, 0x18, 0x0a, 0x07, 0x74, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x18,
-	0x04, 0x20, 0x01, 0x28, 0x0d, 0x52, 0x07, 0x74, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x22, 0x5e,
-	0x0a, 0x10, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x73, 0x75,
-	0x6c, 0x74, 0x12, 0x36, 0x0a, 0x08, 0x73, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x73, 0x18, 0x01,
-	0x20, 0x03, 0x28, 0x0b, 0x32, 0x1a, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54,
-	0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74,
-	0x52, 0x08, 0x73, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x73, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65,
-	0x78, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x22, 0x77,
-	0x0a, 0x11, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x53, 0x65, 0x67, 0x6d,
-	0x65, 0x6e, 0x74, 0x12, 0x0e, 0x0a, 0x02, 0x69, 0x64, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05, 0x52,
-	0x02, 0x69, 0x64, 0x12, 0x14, 0x0a, 0x05, 0x73, 0x74, 0x61, 0x72, 0x74, 0x18, 0x02, 0x20, 0x01,
-	0x28, 0x03, 0x52, 0x05, 0x73, 0x74, 0x61, 0x72, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x65, 0x6e, 0x64,
-	0x18, 0x03, 0x20, 0x01, 0x28, 0x03, 0x52, 0x03, 0x65, 0x6e, 0x64, 0x12, 0x12, 0x0a, 0x04, 0x74,
-	0x65, 0x78, 0x74, 0x18, 0x04, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x12,
-	0x16, 0x0a, 0x06, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73, 0x18, 0x05, 0x20, 0x03, 0x28, 0x05, 0x52,
-	0x06, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73, 0x22, 0xbe, 0x02, 0x0a, 0x14, 0x47, 0x65, 0x6e, 0x65,
-	0x72, 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74,
-	0x12, 0x16, 0x0a, 0x06, 0x68, 0x65, 0x69, 0x67, 0x68, 0x74, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05,
-	0x52, 0x06, 0x68, 0x65, 0x69, 0x67, 0x68, 0x74, 0x12, 0x14, 0x0a, 0x05, 0x77, 0x69, 0x64, 0x74,
-	0x68, 0x18, 0x02, 0x20, 0x01, 0x28, 0x05, 0x52, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, 0x12, 0x12,
-	0x0a, 0x04, 0x6d, 0x6f, 0x64, 0x65, 0x18, 0x03, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x6d, 0x6f,
-	0x64, 0x65, 0x12, 0x12, 0x0a, 0x04, 0x73, 0x74, 0x65, 0x70, 0x18, 0x04, 0x20, 0x01, 0x28, 0x05,
-	0x52, 0x04, 0x73, 0x74, 0x65, 0x70, 0x12, 0x12, 0x0a, 0x04, 0x73, 0x65, 0x65, 0x64, 0x18, 0x05,
-	0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x73, 0x65, 0x65, 0x64, 0x12, 0x27, 0x0a, 0x0f, 0x70, 0x6f,
-	0x73, 0x69, 0x74, 0x69, 0x76, 0x65, 0x5f, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x18, 0x06, 0x20,
-	0x01, 0x28, 0x09, 0x52, 0x0e, 0x70, 0x6f, 0x73, 0x69, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f,
-	0x6d, 0x70, 0x74, 0x12, 0x27, 0x0a, 0x0f, 0x6e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x5f,
-	0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x18, 0x07, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0e, 0x6e, 0x65,
-	0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x12, 0x10, 0x0a, 0x03,
-	0x64, 0x73, 0x74, 0x18, 0x08, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, 0x74, 0x12, 0x10,
-	0x0a, 0x03, 0x73, 0x72, 0x63, 0x18, 0x09, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x73, 0x72, 0x63,
-	0x12, 0x2a, 0x0a, 0x10, 0x45, 0x6e, 0x61, 0x62, 0x6c, 0x65, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x65,
-	0x74, 0x65, 0x72, 0x73, 0x18, 0x0a, 0x20, 0x01, 0x28, 0x09, 0x52, 0x10, 0x45, 0x6e, 0x61, 0x62,
-	0x6c, 0x65, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x65, 0x74, 0x65, 0x72, 0x73, 0x12, 0x1a, 0x0a, 0x08,
-	0x43, 0x4c, 0x49, 0x50, 0x53, 0x6b, 0x69, 0x70, 0x18, 0x0b, 0x20, 0x01, 0x28, 0x05, 0x52, 0x08,
-	0x43, 0x4c, 0x49, 0x50, 0x53, 0x6b, 0x69, 0x70, 0x22, 0x5e, 0x0a, 0x0a, 0x54, 0x54, 0x53, 0x52,
-	0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78, 0x74, 0x18, 0x01,
-	0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x12, 0x14, 0x0a, 0x05, 0x6d, 0x6f,
-	0x64, 0x65, 0x6c, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x05, 0x6d, 0x6f, 0x64, 0x65, 0x6c,
-	0x12, 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x03, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x64,
-	0x73, 0x74, 0x12, 0x14, 0x0a, 0x05, 0x76, 0x6f, 0x69, 0x63, 0x65, 0x18, 0x04, 0x20, 0x01, 0x28,
-	0x09, 0x52, 0x05, 0x76, 0x6f, 0x69, 0x63, 0x65, 0x22, 0x46, 0x0a, 0x14, 0x54, 0x6f, 0x6b, 0x65,
-	0x6e, 0x69, 0x7a, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65,
-	0x12, 0x16, 0x0a, 0x06, 0x6c, 0x65, 0x6e, 0x67, 0x74, 0x68, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05,
-	0x52, 0x06, 0x6c, 0x65, 0x6e, 0x67, 0x74, 0x68, 0x12, 0x16, 0x0a, 0x06, 0x74, 0x6f, 0x6b, 0x65,
-	0x6e, 0x73, 0x18, 0x02, 0x20, 0x03, 0x28, 0x05, 0x52, 0x06, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73,
-	0x22, 0xac, 0x01, 0x0a, 0x0f, 0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x55, 0x73, 0x61, 0x67, 0x65,
-	0x44, 0x61, 0x74, 0x61, 0x12, 0x14, 0x0a, 0x05, 0x74, 0x6f, 0x74, 0x61, 0x6c, 0x18, 0x01, 0x20,
-	0x01, 0x28, 0x04, 0x52, 0x05, 0x74, 0x6f, 0x74, 0x61, 0x6c, 0x12, 0x45, 0x0a, 0x09, 0x62, 0x72,
-	0x65, 0x61, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x18, 0x02, 0x20, 0x03, 0x28, 0x0b, 0x32, 0x27, 0x2e,
-	0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x55, 0x73,
-	0x61, 0x67, 0x65, 0x44, 0x61, 0x74, 0x61, 0x2e, 0x42, 0x72, 0x65, 0x61, 0x6b, 0x64, 0x6f, 0x77,
-	0x6e, 0x45, 0x6e, 0x74, 0x72, 0x79, 0x52, 0x09, 0x62, 0x72, 0x65, 0x61, 0x6b, 0x64, 0x6f, 0x77,
-	0x6e, 0x1a, 0x3c, 0x0a, 0x0e, 0x42, 0x72, 0x65, 0x61, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x45, 0x6e,
-	0x74, 0x72, 0x79, 0x12, 0x10, 0x0a, 0x03, 0x6b, 0x65, 0x79, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09,
-	0x52, 0x03, 0x6b, 0x65, 0x79, 0x12, 0x14, 0x0a, 0x05, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x18, 0x02,
-	0x20, 0x01, 0x28, 0x04, 0x52, 0x05, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3a, 0x02, 0x38, 0x01, 0x22,
-	0xbc, 0x01, 0x0a, 0x0e, 0x53, 0x74, 0x61, 0x74, 0x75, 0x73, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e,
-	0x73, 0x65, 0x12, 0x33, 0x0a, 0x05, 0x73, 0x74, 0x61, 0x74, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28,
-	0x0e, 0x32, 0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x61, 0x74,
-	0x75, 0x73, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x2e, 0x53, 0x74, 0x61, 0x74, 0x65,
-	0x52, 0x05, 0x73, 0x74, 0x61, 0x74, 0x65, 0x12, 0x30, 0x0a, 0x06, 0x6d, 0x65, 0x6d, 0x6f, 0x72,
-	0x79, 0x18, 0x02, 0x20, 0x01, 0x28, 0x0b, 0x32, 0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e,
-	0x64, 0x2e, 0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x55, 0x73, 0x61, 0x67, 0x65, 0x44, 0x61, 0x74,
-	0x61, 0x52, 0x06, 0x6d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x22, 0x43, 0x0a, 0x05, 0x53, 0x74, 0x61,
-	0x74, 0x65, 0x12, 0x11, 0x0a, 0x0d, 0x55, 0x4e, 0x49, 0x4e, 0x49, 0x54, 0x49, 0x41, 0x4c, 0x49,
-	0x5a, 0x45, 0x44, 0x10, 0x00, 0x12, 0x08, 0x0a, 0x04, 0x42, 0x55, 0x53, 0x59, 0x10, 0x01, 0x12,
-	0x09, 0x0a, 0x05, 0x52, 0x45, 0x41, 0x44, 0x59, 0x10, 0x02, 0x12, 0x12, 0x0a, 0x05, 0x45, 0x52,
-	0x52, 0x4f, 0x52, 0x10, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x01, 0x22, 0x37,
-	0x0a, 0x07, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x12, 0x12, 0x0a, 0x04, 0x72, 0x6f, 0x6c,
-	0x65, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x72, 0x6f, 0x6c, 0x65, 0x12, 0x18, 0x0a,
-	0x07, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x07,
-	0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x32, 0xfb, 0x06, 0x0a, 0x07, 0x42, 0x61, 0x63, 0x6b,
-	0x65, 0x6e, 0x64, 0x12, 0x32, 0x0a, 0x06, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x12, 0x16, 0x2e,
-	0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65,
-	0x73, 0x73, 0x61, 0x67, 0x65, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e,
-	0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x34, 0x0a, 0x07, 0x50, 0x72, 0x65, 0x64, 0x69,
-	0x63, 0x74, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65,
-	0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62, 0x61,
-	0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x35, 0x0a,
-	0x09, 0x4c, 0x6f, 0x61, 0x64, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x15, 0x2e, 0x62, 0x61, 0x63,
-	0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e,
-	0x73, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75,
-	0x6c, 0x74, 0x22, 0x00, 0x12, 0x3c, 0x0a, 0x0d, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x53,
-	0x74, 0x72, 0x65, 0x61, 0x6d, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e,
-	0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e,
-	0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00,
-	0x30, 0x01, 0x12, 0x40, 0x0a, 0x09, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x12,
-	0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63,
-	0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65,
-	0x6e, 0x64, 0x2e, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75,
-	0x6c, 0x74, 0x22, 0x00, 0x12, 0x41, 0x0a, 0x0d, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65,
-	0x49, 0x6d, 0x61, 0x67, 0x65, 0x12, 0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e,
-	0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x52, 0x65, 0x71,
-	0x75, 0x65, 0x73, 0x74, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52,
-	0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x4d, 0x0a, 0x12, 0x41, 0x75, 0x64, 0x69, 0x6f,
-	0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x12, 0x1a, 0x2e,
-	0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69,
-	0x70, 0x74, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x19, 0x2e, 0x62, 0x61, 0x63, 0x6b,
-	0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65,
-	0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x2d, 0x0a, 0x03, 0x54, 0x54, 0x53, 0x12, 0x13, 0x2e,
-	0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65,
-	0x73, 0x74, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73,
-	0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x4a, 0x0a, 0x0e, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x69, 0x7a,
-	0x65, 0x53, 0x74, 0x72, 0x69, 0x6e, 0x67, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e,
-	0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73,
-	0x1a, 0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x6f, 0x6b, 0x65, 0x6e,
-	0x69, 0x7a, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x22,
-	0x00, 0x12, 0x3b, 0x0a, 0x06, 0x53, 0x74, 0x61, 0x74, 0x75, 0x73, 0x12, 0x16, 0x2e, 0x62, 0x61,
-	0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65, 0x73, 0x73,
-	0x61, 0x67, 0x65, 0x1a, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74,
-	0x61, 0x74, 0x75, 0x73, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x22, 0x00, 0x12, 0x39,
-	0x0a, 0x09, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x53, 0x65, 0x74, 0x12, 0x19, 0x2e, 0x62, 0x61,
-	0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x53, 0x65, 0x74, 0x4f,
-	0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64,
-	0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x3f, 0x0a, 0x0c, 0x53, 0x74, 0x6f,
-	0x72, 0x65, 0x73, 0x44, 0x65, 0x6c, 0x65, 0x74, 0x65, 0x12, 0x1c, 0x2e, 0x62, 0x61, 0x63, 0x6b,
-	0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x44, 0x65, 0x6c, 0x65, 0x74, 0x65,
-	0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e,
-	0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x42, 0x0a, 0x09, 0x53, 0x74,
-	0x6f, 0x72, 0x65, 0x73, 0x47, 0x65, 0x74, 0x12, 0x19, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e,
-	0x64, 0x2e, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x47, 0x65, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f,
-	0x6e, 0x73, 0x1a, 0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f,
-	0x72, 0x65, 0x73, 0x47, 0x65, 0x74, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x45,
-	0x0a, 0x0a, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x46, 0x69, 0x6e, 0x64, 0x12, 0x1a, 0x2e, 0x62,
-	0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x46, 0x69, 0x6e,
-	0x64, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x19, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65,
-	0x6e, 0x64, 0x2e, 0x53, 0x74, 0x6f, 0x72, 0x65, 0x73, 0x46, 0x69, 0x6e, 0x64, 0x52, 0x65, 0x73,
-	0x75, 0x6c, 0x74, 0x22, 0x00, 0x42, 0x5a, 0x0a, 0x19, 0x69, 0x6f, 0x2e, 0x73, 0x6b, 0x79, 0x6e,
-	0x65, 0x74, 0x2e, 0x6c, 0x6f, 0x63, 0x61, 0x6c, 0x61, 0x69, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65,
-	0x6e, 0x64, 0x42, 0x0e, 0x4c, 0x6f, 0x63, 0x61, 0x6c, 0x41, 0x49, 0x42, 0x61, 0x63, 0x6b, 0x65,
-	0x6e, 0x64, 0x50, 0x01, 0x5a, 0x2b, 0x67, 0x69, 0x74, 0x68, 0x75, 0x62, 0x2e, 0x63, 0x6f, 0x6d,
-	0x2f, 0x67, 0x6f, 0x2d, 0x73, 0x6b, 0x79, 0x6e, 0x65, 0x74, 0x2f, 0x4c, 0x6f, 0x63, 0x61, 0x6c,
-	0x41, 0x49, 0x2f, 0x70, 0x6b, 0x67, 0x2f, 0x67, 0x72, 0x70, 0x63, 0x2f, 0x70, 0x72, 0x6f, 0x74,
-	0x6f, 0x62, 0x06, 0x70, 0x72, 0x6f, 0x74, 0x6f, 0x33,
-}
-
-var (
-	file_backend_proto_rawDescOnce sync.Once
-	file_backend_proto_rawDescData = file_backend_proto_rawDesc
-)
-
-func file_backend_proto_rawDescGZIP() []byte {
-	file_backend_proto_rawDescOnce.Do(func() {
-		file_backend_proto_rawDescData = protoimpl.X.CompressGZIP(file_backend_proto_rawDescData)
-	})
-	return file_backend_proto_rawDescData
-}
-
-var file_backend_proto_enumTypes = make([]protoimpl.EnumInfo, 1)
-var file_backend_proto_msgTypes = make([]protoimpl.MessageInfo, 24)
-var file_backend_proto_goTypes = []interface{}{
-	(StatusResponse_State)(0),    // 0: backend.StatusResponse.State
-	(*StoresKey)(nil),            // 1: backend.StoresKey
-	(*StoresValue)(nil),          // 2: backend.StoresValue
-	(*StoresSetOptions)(nil),     // 3: backend.StoresSetOptions
-	(*StoresDeleteOptions)(nil),  // 4: backend.StoresDeleteOptions
-	(*StoresGetOptions)(nil),     // 5: backend.StoresGetOptions
-	(*StoresGetResult)(nil),      // 6: backend.StoresGetResult
-	(*StoresFindOptions)(nil),    // 7: backend.StoresFindOptions
-	(*StoresFindResult)(nil),     // 8: backend.StoresFindResult
-	(*HealthMessage)(nil),        // 9: backend.HealthMessage
-	(*PredictOptions)(nil),       // 10: backend.PredictOptions
-	(*Reply)(nil),                // 11: backend.Reply
-	(*ModelOptions)(nil),         // 12: backend.ModelOptions
-	(*Result)(nil),               // 13: backend.Result
-	(*EmbeddingResult)(nil),      // 14: backend.EmbeddingResult
-	(*TranscriptRequest)(nil),    // 15: backend.TranscriptRequest
-	(*TranscriptResult)(nil),     // 16: backend.TranscriptResult
-	(*TranscriptSegment)(nil),    // 17: backend.TranscriptSegment
-	(*GenerateImageRequest)(nil), // 18: backend.GenerateImageRequest
-	(*TTSRequest)(nil),           // 19: backend.TTSRequest
-	(*TokenizationResponse)(nil), // 20: backend.TokenizationResponse
-	(*MemoryUsageData)(nil),      // 21: backend.MemoryUsageData
-	(*StatusResponse)(nil),       // 22: backend.StatusResponse
-	(*Message)(nil),              // 23: backend.Message
-	nil,                          // 24: backend.MemoryUsageData.BreakdownEntry
-}
-var file_backend_proto_depIdxs = []int32{
-	1,  // 0: backend.StoresSetOptions.Keys:type_name -> backend.StoresKey
-	2,  // 1: backend.StoresSetOptions.Values:type_name -> backend.StoresValue
-	1,  // 2: backend.StoresDeleteOptions.Keys:type_name -> backend.StoresKey
-	1,  // 3: backend.StoresGetOptions.Keys:type_name -> backend.StoresKey
-	1,  // 4: backend.StoresGetResult.Keys:type_name -> backend.StoresKey
-	2,  // 5: backend.StoresGetResult.Values:type_name -> backend.StoresValue
-	1,  // 6: backend.StoresFindOptions.Key:type_name -> backend.StoresKey
-	1,  // 7: backend.StoresFindResult.Keys:type_name -> backend.StoresKey
-	2,  // 8: backend.StoresFindResult.Values:type_name -> backend.StoresValue
-	23, // 9: backend.PredictOptions.Messages:type_name -> backend.Message
-	17, // 10: backend.TranscriptResult.segments:type_name -> backend.TranscriptSegment
-	24, // 11: backend.MemoryUsageData.breakdown:type_name -> backend.MemoryUsageData.BreakdownEntry
-	0,  // 12: backend.StatusResponse.state:type_name -> backend.StatusResponse.State
-	21, // 13: backend.StatusResponse.memory:type_name -> backend.MemoryUsageData
-	9,  // 14: backend.Backend.Health:input_type -> backend.HealthMessage
-	10, // 15: backend.Backend.Predict:input_type -> backend.PredictOptions
-	12, // 16: backend.Backend.LoadModel:input_type -> backend.ModelOptions
-	10, // 17: backend.Backend.PredictStream:input_type -> backend.PredictOptions
-	10, // 18: backend.Backend.Embedding:input_type -> backend.PredictOptions
-	18, // 19: backend.Backend.GenerateImage:input_type -> backend.GenerateImageRequest
-	15, // 20: backend.Backend.AudioTranscription:input_type -> backend.TranscriptRequest
-	19, // 21: backend.Backend.TTS:input_type -> backend.TTSRequest
-	10, // 22: backend.Backend.TokenizeString:input_type -> backend.PredictOptions
-	9,  // 23: backend.Backend.Status:input_type -> backend.HealthMessage
-	3,  // 24: backend.Backend.StoresSet:input_type -> backend.StoresSetOptions
-	4,  // 25: backend.Backend.StoresDelete:input_type -> backend.StoresDeleteOptions
-	5,  // 26: backend.Backend.StoresGet:input_type -> backend.StoresGetOptions
-	7,  // 27: backend.Backend.StoresFind:input_type -> backend.StoresFindOptions
-	11, // 28: backend.Backend.Health:output_type -> backend.Reply
-	11, // 29: backend.Backend.Predict:output_type -> backend.Reply
-	13, // 30: backend.Backend.LoadModel:output_type -> backend.Result
-	11, // 31: backend.Backend.PredictStream:output_type -> backend.Reply
-	14, // 32: backend.Backend.Embedding:output_type -> backend.EmbeddingResult
-	13, // 33: backend.Backend.GenerateImage:output_type -> backend.Result
-	16, // 34: backend.Backend.AudioTranscription:output_type -> backend.TranscriptResult
-	13, // 35: backend.Backend.TTS:output_type -> backend.Result
-	20, // 36: backend.Backend.TokenizeString:output_type -> backend.TokenizationResponse
-	22, // 37: backend.Backend.Status:output_type -> backend.StatusResponse
-	13, // 38: backend.Backend.StoresSet:output_type -> backend.Result
-	13, // 39: backend.Backend.StoresDelete:output_type -> backend.Result
-	6,  // 40: backend.Backend.StoresGet:output_type -> backend.StoresGetResult
-	8,  // 41: backend.Backend.StoresFind:output_type -> backend.StoresFindResult
-	28, // [28:42] is the sub-list for method output_type
-	14, // [14:28] is the sub-list for method input_type
-	14, // [14:14] is the sub-list for extension type_name
-	14, // [14:14] is the sub-list for extension extendee
-	0,  // [0:14] is the sub-list for field type_name
-}
-
-func init() { file_backend_proto_init() }
-func file_backend_proto_init() {
-	if File_backend_proto != nil {
-		return
-	}
-	if !protoimpl.UnsafeEnabled {
-		file_backend_proto_msgTypes[0].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*StoresKey); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[1].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*StoresValue); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[2].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*StoresSetOptions); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[3].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*StoresDeleteOptions); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[4].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*StoresGetOptions); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[5].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*StoresGetResult); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[6].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*StoresFindOptions); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[7].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*StoresFindResult); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[8].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*HealthMessage); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[9].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*PredictOptions); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[10].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*Reply); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[11].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*ModelOptions); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[12].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*Result); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[13].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*EmbeddingResult); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[14].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*TranscriptRequest); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[15].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*TranscriptResult); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[16].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*TranscriptSegment); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[17].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*GenerateImageRequest); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[18].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*TTSRequest); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[19].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*TokenizationResponse); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[20].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*MemoryUsageData); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[21].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*StatusResponse); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-		file_backend_proto_msgTypes[22].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*Message); i {
-			case 0:
-				return &v.state
-			case 1:
-				return &v.sizeCache
-			case 2:
-				return &v.unknownFields
-			default:
-				return nil
-			}
-		}
-	}
-	type x struct{}
-	out := protoimpl.TypeBuilder{
-		File: protoimpl.DescBuilder{
-			GoPackagePath: reflect.TypeOf(x{}).PkgPath(),
-			RawDescriptor: file_backend_proto_rawDesc,
-			NumEnums:      1,
-			NumMessages:   24,
-			NumExtensions: 0,
-			NumServices:   1,
-		},
-		GoTypes:           file_backend_proto_goTypes,
-		DependencyIndexes: file_backend_proto_depIdxs,
-		EnumInfos:         file_backend_proto_enumTypes,
-		MessageInfos:      file_backend_proto_msgTypes,
-	}.Build()
-	File_backend_proto = out.File
-	file_backend_proto_rawDesc = nil
-	file_backend_proto_goTypes = nil
-	file_backend_proto_depIdxs = nil
-}
diff --git a/pkg/grpc/proto/backend_grpc.pb.go b/pkg/grpc/proto/backend_grpc.pb.go
deleted file mode 100644
index a1f442e0..00000000
--- a/pkg/grpc/proto/backend_grpc.pb.go
+++ /dev/null
@@ -1,618 +0,0 @@
-// Code generated by protoc-gen-go-grpc. DO NOT EDIT.
-// versions:
-// - protoc-gen-go-grpc v1.3.0
-// - protoc             v5.26.1
-// source: backend.proto
-
-package proto
-
-import (
-	context "context"
-	grpc "google.golang.org/grpc"
-	codes "google.golang.org/grpc/codes"
-	status "google.golang.org/grpc/status"
-)
-
-// This is a compile-time assertion to ensure that this generated file
-// is compatible with the grpc package it is being compiled against.
-// Requires gRPC-Go v1.32.0 or later.
-const _ = grpc.SupportPackageIsVersion7
-
-const (
-	Backend_Health_FullMethodName             = "/backend.Backend/Health"
-	Backend_Predict_FullMethodName            = "/backend.Backend/Predict"
-	Backend_LoadModel_FullMethodName          = "/backend.Backend/LoadModel"
-	Backend_PredictStream_FullMethodName      = "/backend.Backend/PredictStream"
-	Backend_Embedding_FullMethodName          = "/backend.Backend/Embedding"
-	Backend_GenerateImage_FullMethodName      = "/backend.Backend/GenerateImage"
-	Backend_AudioTranscription_FullMethodName = "/backend.Backend/AudioTranscription"
-	Backend_TTS_FullMethodName                = "/backend.Backend/TTS"
-	Backend_TokenizeString_FullMethodName     = "/backend.Backend/TokenizeString"
-	Backend_Status_FullMethodName             = "/backend.Backend/Status"
-	Backend_StoresSet_FullMethodName          = "/backend.Backend/StoresSet"
-	Backend_StoresDelete_FullMethodName       = "/backend.Backend/StoresDelete"
-	Backend_StoresGet_FullMethodName          = "/backend.Backend/StoresGet"
-	Backend_StoresFind_FullMethodName         = "/backend.Backend/StoresFind"
-)
-
-// BackendClient is the client API for Backend service.
-//
-// For semantics around ctx use and closing/ending streaming RPCs, please refer to https://pkg.go.dev/google.golang.org/grpc/?tab=doc#ClientConn.NewStream.
-type BackendClient interface {
-	Health(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*Reply, error)
-	Predict(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*Reply, error)
-	LoadModel(ctx context.Context, in *ModelOptions, opts ...grpc.CallOption) (*Result, error)
-	PredictStream(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (Backend_PredictStreamClient, error)
-	Embedding(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*EmbeddingResult, error)
-	GenerateImage(ctx context.Context, in *GenerateImageRequest, opts ...grpc.CallOption) (*Result, error)
-	AudioTranscription(ctx context.Context, in *TranscriptRequest, opts ...grpc.CallOption) (*TranscriptResult, error)
-	TTS(ctx context.Context, in *TTSRequest, opts ...grpc.CallOption) (*Result, error)
-	TokenizeString(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*TokenizationResponse, error)
-	Status(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*StatusResponse, error)
-	StoresSet(ctx context.Context, in *StoresSetOptions, opts ...grpc.CallOption) (*Result, error)
-	StoresDelete(ctx context.Context, in *StoresDeleteOptions, opts ...grpc.CallOption) (*Result, error)
-	StoresGet(ctx context.Context, in *StoresGetOptions, opts ...grpc.CallOption) (*StoresGetResult, error)
-	StoresFind(ctx context.Context, in *StoresFindOptions, opts ...grpc.CallOption) (*StoresFindResult, error)
-}
-
-type backendClient struct {
-	cc grpc.ClientConnInterface
-}
-
-func NewBackendClient(cc grpc.ClientConnInterface) BackendClient {
-	return &backendClient{cc}
-}
-
-func (c *backendClient) Health(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*Reply, error) {
-	out := new(Reply)
-	err := c.cc.Invoke(ctx, Backend_Health_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) Predict(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*Reply, error) {
-	out := new(Reply)
-	err := c.cc.Invoke(ctx, Backend_Predict_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) LoadModel(ctx context.Context, in *ModelOptions, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, Backend_LoadModel_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) PredictStream(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (Backend_PredictStreamClient, error) {
-	stream, err := c.cc.NewStream(ctx, &Backend_ServiceDesc.Streams[0], Backend_PredictStream_FullMethodName, opts...)
-	if err != nil {
-		return nil, err
-	}
-	x := &backendPredictStreamClient{stream}
-	if err := x.ClientStream.SendMsg(in); err != nil {
-		return nil, err
-	}
-	if err := x.ClientStream.CloseSend(); err != nil {
-		return nil, err
-	}
-	return x, nil
-}
-
-type Backend_PredictStreamClient interface {
-	Recv() (*Reply, error)
-	grpc.ClientStream
-}
-
-type backendPredictStreamClient struct {
-	grpc.ClientStream
-}
-
-func (x *backendPredictStreamClient) Recv() (*Reply, error) {
-	m := new(Reply)
-	if err := x.ClientStream.RecvMsg(m); err != nil {
-		return nil, err
-	}
-	return m, nil
-}
-
-func (c *backendClient) Embedding(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*EmbeddingResult, error) {
-	out := new(EmbeddingResult)
-	err := c.cc.Invoke(ctx, Backend_Embedding_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) GenerateImage(ctx context.Context, in *GenerateImageRequest, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, Backend_GenerateImage_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) AudioTranscription(ctx context.Context, in *TranscriptRequest, opts ...grpc.CallOption) (*TranscriptResult, error) {
-	out := new(TranscriptResult)
-	err := c.cc.Invoke(ctx, Backend_AudioTranscription_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) TTS(ctx context.Context, in *TTSRequest, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, Backend_TTS_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) TokenizeString(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*TokenizationResponse, error) {
-	out := new(TokenizationResponse)
-	err := c.cc.Invoke(ctx, Backend_TokenizeString_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) Status(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*StatusResponse, error) {
-	out := new(StatusResponse)
-	err := c.cc.Invoke(ctx, Backend_Status_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) StoresSet(ctx context.Context, in *StoresSetOptions, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, Backend_StoresSet_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) StoresDelete(ctx context.Context, in *StoresDeleteOptions, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, Backend_StoresDelete_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) StoresGet(ctx context.Context, in *StoresGetOptions, opts ...grpc.CallOption) (*StoresGetResult, error) {
-	out := new(StoresGetResult)
-	err := c.cc.Invoke(ctx, Backend_StoresGet_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) StoresFind(ctx context.Context, in *StoresFindOptions, opts ...grpc.CallOption) (*StoresFindResult, error) {
-	out := new(StoresFindResult)
-	err := c.cc.Invoke(ctx, Backend_StoresFind_FullMethodName, in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-// BackendServer is the server API for Backend service.
-// All implementations must embed UnimplementedBackendServer
-// for forward compatibility
-type BackendServer interface {
-	Health(context.Context, *HealthMessage) (*Reply, error)
-	Predict(context.Context, *PredictOptions) (*Reply, error)
-	LoadModel(context.Context, *ModelOptions) (*Result, error)
-	PredictStream(*PredictOptions, Backend_PredictStreamServer) error
-	Embedding(context.Context, *PredictOptions) (*EmbeddingResult, error)
-	GenerateImage(context.Context, *GenerateImageRequest) (*Result, error)
-	AudioTranscription(context.Context, *TranscriptRequest) (*TranscriptResult, error)
-	TTS(context.Context, *TTSRequest) (*Result, error)
-	TokenizeString(context.Context, *PredictOptions) (*TokenizationResponse, error)
-	Status(context.Context, *HealthMessage) (*StatusResponse, error)
-	StoresSet(context.Context, *StoresSetOptions) (*Result, error)
-	StoresDelete(context.Context, *StoresDeleteOptions) (*Result, error)
-	StoresGet(context.Context, *StoresGetOptions) (*StoresGetResult, error)
-	StoresFind(context.Context, *StoresFindOptions) (*StoresFindResult, error)
-	mustEmbedUnimplementedBackendServer()
-}
-
-// UnimplementedBackendServer must be embedded to have forward compatible implementations.
-type UnimplementedBackendServer struct {
-}
-
-func (UnimplementedBackendServer) Health(context.Context, *HealthMessage) (*Reply, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Health not implemented")
-}
-func (UnimplementedBackendServer) Predict(context.Context, *PredictOptions) (*Reply, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Predict not implemented")
-}
-func (UnimplementedBackendServer) LoadModel(context.Context, *ModelOptions) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method LoadModel not implemented")
-}
-func (UnimplementedBackendServer) PredictStream(*PredictOptions, Backend_PredictStreamServer) error {
-	return status.Errorf(codes.Unimplemented, "method PredictStream not implemented")
-}
-func (UnimplementedBackendServer) Embedding(context.Context, *PredictOptions) (*EmbeddingResult, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Embedding not implemented")
-}
-func (UnimplementedBackendServer) GenerateImage(context.Context, *GenerateImageRequest) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method GenerateImage not implemented")
-}
-func (UnimplementedBackendServer) AudioTranscription(context.Context, *TranscriptRequest) (*TranscriptResult, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method AudioTranscription not implemented")
-}
-func (UnimplementedBackendServer) TTS(context.Context, *TTSRequest) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method TTS not implemented")
-}
-func (UnimplementedBackendServer) TokenizeString(context.Context, *PredictOptions) (*TokenizationResponse, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method TokenizeString not implemented")
-}
-func (UnimplementedBackendServer) Status(context.Context, *HealthMessage) (*StatusResponse, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Status not implemented")
-}
-func (UnimplementedBackendServer) StoresSet(context.Context, *StoresSetOptions) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method StoresSet not implemented")
-}
-func (UnimplementedBackendServer) StoresDelete(context.Context, *StoresDeleteOptions) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method StoresDelete not implemented")
-}
-func (UnimplementedBackendServer) StoresGet(context.Context, *StoresGetOptions) (*StoresGetResult, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method StoresGet not implemented")
-}
-func (UnimplementedBackendServer) StoresFind(context.Context, *StoresFindOptions) (*StoresFindResult, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method StoresFind not implemented")
-}
-func (UnimplementedBackendServer) mustEmbedUnimplementedBackendServer() {}
-
-// UnsafeBackendServer may be embedded to opt out of forward compatibility for this service.
-// Use of this interface is not recommended, as added methods to BackendServer will
-// result in compilation errors.
-type UnsafeBackendServer interface {
-	mustEmbedUnimplementedBackendServer()
-}
-
-func RegisterBackendServer(s grpc.ServiceRegistrar, srv BackendServer) {
-	s.RegisterService(&Backend_ServiceDesc, srv)
-}
-
-func _Backend_Health_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(HealthMessage)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Health(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_Health_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Health(ctx, req.(*HealthMessage))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_Predict_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(PredictOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Predict(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_Predict_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Predict(ctx, req.(*PredictOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_LoadModel_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(ModelOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).LoadModel(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_LoadModel_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).LoadModel(ctx, req.(*ModelOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_PredictStream_Handler(srv interface{}, stream grpc.ServerStream) error {
-	m := new(PredictOptions)
-	if err := stream.RecvMsg(m); err != nil {
-		return err
-	}
-	return srv.(BackendServer).PredictStream(m, &backendPredictStreamServer{stream})
-}
-
-type Backend_PredictStreamServer interface {
-	Send(*Reply) error
-	grpc.ServerStream
-}
-
-type backendPredictStreamServer struct {
-	grpc.ServerStream
-}
-
-func (x *backendPredictStreamServer) Send(m *Reply) error {
-	return x.ServerStream.SendMsg(m)
-}
-
-func _Backend_Embedding_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(PredictOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Embedding(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_Embedding_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Embedding(ctx, req.(*PredictOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_GenerateImage_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(GenerateImageRequest)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).GenerateImage(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_GenerateImage_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).GenerateImage(ctx, req.(*GenerateImageRequest))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_AudioTranscription_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(TranscriptRequest)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).AudioTranscription(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_AudioTranscription_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).AudioTranscription(ctx, req.(*TranscriptRequest))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_TTS_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(TTSRequest)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).TTS(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_TTS_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).TTS(ctx, req.(*TTSRequest))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_TokenizeString_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(PredictOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).TokenizeString(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_TokenizeString_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).TokenizeString(ctx, req.(*PredictOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_Status_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(HealthMessage)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Status(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_Status_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Status(ctx, req.(*HealthMessage))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_StoresSet_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(StoresSetOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).StoresSet(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_StoresSet_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).StoresSet(ctx, req.(*StoresSetOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_StoresDelete_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(StoresDeleteOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).StoresDelete(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_StoresDelete_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).StoresDelete(ctx, req.(*StoresDeleteOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_StoresGet_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(StoresGetOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).StoresGet(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_StoresGet_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).StoresGet(ctx, req.(*StoresGetOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_StoresFind_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(StoresFindOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).StoresFind(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: Backend_StoresFind_FullMethodName,
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).StoresFind(ctx, req.(*StoresFindOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-// Backend_ServiceDesc is the grpc.ServiceDesc for Backend service.
-// It's only intended for direct use with grpc.RegisterService,
-// and not to be introspected or modified (even as a copy)
-var Backend_ServiceDesc = grpc.ServiceDesc{
-	ServiceName: "backend.Backend",
-	HandlerType: (*BackendServer)(nil),
-	Methods: []grpc.MethodDesc{
-		{
-			MethodName: "Health",
-			Handler:    _Backend_Health_Handler,
-		},
-		{
-			MethodName: "Predict",
-			Handler:    _Backend_Predict_Handler,
-		},
-		{
-			MethodName: "LoadModel",
-			Handler:    _Backend_LoadModel_Handler,
-		},
-		{
-			MethodName: "Embedding",
-			Handler:    _Backend_Embedding_Handler,
-		},
-		{
-			MethodName: "GenerateImage",
-			Handler:    _Backend_GenerateImage_Handler,
-		},
-		{
-			MethodName: "AudioTranscription",
-			Handler:    _Backend_AudioTranscription_Handler,
-		},
-		{
-			MethodName: "TTS",
-			Handler:    _Backend_TTS_Handler,
-		},
-		{
-			MethodName: "TokenizeString",
-			Handler:    _Backend_TokenizeString_Handler,
-		},
-		{
-			MethodName: "Status",
-			Handler:    _Backend_Status_Handler,
-		},
-		{
-			MethodName: "StoresSet",
-			Handler:    _Backend_StoresSet_Handler,
-		},
-		{
-			MethodName: "StoresDelete",
-			Handler:    _Backend_StoresDelete_Handler,
-		},
-		{
-			MethodName: "StoresGet",
-			Handler:    _Backend_StoresGet_Handler,
-		},
-		{
-			MethodName: "StoresFind",
-			Handler:    _Backend_StoresFind_Handler,
-		},
-	},
-	Streams: []grpc.StreamDesc{
-		{
-			StreamName:    "PredictStream",
-			Handler:       _Backend_PredictStream_Handler,
-			ServerStreams: true,
-		},
-	},
-	Metadata: "backend.proto",
-}