working elf server

lollms/server/.gitignore

@@ -0,0 +1 @@
+elf_global_paths_cfg.yaml

lollms/server/configs/.gitignore

@@ -0,0 +1 @@
+local_config.yaml

lollms/server/configs/config.yaml

@@ -0,0 +1,96 @@
+# =================== Lord Of Large Language Models Configuration file =========================== 
+version: 39
+binding_name: null
+model_name: null
+
+
+
+# Host information
+host: localhost
+port: 9601
+
+# Genreration parameters 
+discussion_prompt_separator: "!@>"
+seed: -1
+n_predict: 1024
+ctx_size: 4084
+min_n_predict: 512
+temperature: 0.9
+top_k: 50
+top_p: 0.95
+repeat_last_n: 40
+repeat_penalty: 1.2
+
+n_threads: 8
+
+#Personality parameters
+personalities: ["generic/lollms"]
+active_personality_id: 0
+override_personality_model_parameters: false #if true the personality parameters are overriden by those of the configuration (may affect personality behaviour) 
+
+extensions: []
+
+user_name: user
+user_description: ""
+use_user_name_in_discussions: false
+user_avatar: default_user.svg
+use_user_informations_in_discussion: false
+
+# UI parameters
+db_path: database.db
+
+# Automatic updates
+debug: False
+auto_update: true
+auto_save: true
+auto_title: false
+# Enables gpu usage
+enable_gpu: true
+# Automatically open the browser
+auto_show_browser: true
+
+# Voice service
+enable_voice_service: false
+xtts_base_url: http://127.0.0.1:8020
+auto_read: false
+current_voice: null
+current_language: en
+
+# Image generation service
+enable_sd_service: false
+sd_base_url: http://127.0.0.1:7860
+
+# Audio
+media_on: false
+audio_in_language: 'en-US'
+auto_speak: false
+audio_out_voice: null
+audio_pitch: 1
+audio_auto_send_input: true
+audio_silenceTimer: 5000
+
+# Data vectorization
+use_discussions_history: false # Activate vectorizing previous conversations
+summerize_discussion: false # activate discussion summary (better but adds computation time)
+max_summary_size: 512 # in tokens
+data_vectorization_visualize_on_vectorization: false
+use_files: true # Activate using files
+data_vectorization_activate: true # To activate/deactivate data vectorization
+data_vectorization_method: "tfidf_vectorizer" #"model_embedding" or "tfidf_vectorizer"
+data_visualization_method: "PCA" #"PCA" or "TSNE"
+data_vectorization_save_db: False # For each new session, new files
+data_vectorization_chunk_size: 512 # chunk size
+data_vectorization_overlap_size: 128 # overlap between chunks size
+data_vectorization_nb_chunks: 2 # number of chunks to use
+data_vectorization_build_keys_words: false # If true, when querrying the database, we use keywords generated from the user prompt instead of the prompt itself.
+data_vectorization_force_first_chunk: false # If true, the first chunk of the document will systematically be used
+data_vectorization_make_persistance: false # If true, the data will be persistant webween runs
+
+
+# Helpers
+pdf_latex_path: null
+
+# boosting information
+positive_boost: null
+negative_boost: null
+force_output_language_to_be: null

lollms/server/elf_server.py

@@ -19,6 +19,8 @@ class LOLLMSElfServer(LollmsApplication):
         lollms_paths: LollmsPaths,
         load_binding=True,
         load_model=True,
+        load_voice_service=True,
+        load_sd_service=True,
         try_select_binding=False,
         try_select_model=False,
         callback=None,
@@ -30,6 +32,8 @@ class LOLLMSElfServer(LollmsApplication):
                 lollms_paths,
                 load_binding=load_binding,
                 load_model=load_model,
+                load_sd_service=load_sd_service,
+                load_voice_service=load_voice_service,
                 try_select_binding=try_select_binding,
                 try_select_model=try_select_model,
                 callback=callback,
@@ -46,6 +50,8 @@ class LOLLMSElfServer(LollmsApplication):
         lollms_paths: LollmsPaths,
         load_binding=True,
         load_model=True,
+        load_voice_service=True,
+        load_sd_service=True,
         try_select_binding=False,
         try_select_model=False,
         callback=None,
@@ -57,6 +63,8 @@ class LOLLMSElfServer(LollmsApplication):
             lollms_paths,
             load_binding=load_binding,
             load_model=load_model,
+            load_sd_service=load_sd_service,
+            load_voice_service=load_voice_service,
             try_select_binding=try_select_binding,
             try_select_model=try_select_model,
             callback=callback,

lollms/server/endpoints/lollms_generator.py

@@ -0,0 +1,34 @@
+from fastapi import APIRouter
+from lollms.server.elf_server import LOLLMSElfServer
+from pydantic import BaseModel
+from starlette.responses import StreamingResponse
+
+class GenerateRequest(BaseModel):
+    text: str
+    n_predict: int = 1024
+    stream: bool = False
+
+router = APIRouter()
+elf_server = LOLLMSElfServer.get_instance()
+
+@router.post("/generate")
+def generate(request_data: GenerateRequest):
+    text = request_data.text
+    n_predict = request_data.n_predict
+    stream = request_data.stream
+    
+    if elf_server.binding is not None:
+        if stream:
+            def generate_chunks():
+                def callback(chunk):
+                    # Yield each chunk of data
+                    yield chunk
+                
+                elf_server.binding.generate(text, n_predict, callback=callback)
+            
+            return StreamingResponse(generate_chunks())
+        else:
+            output = elf_server.binding.generate(text, n_predict)
+            return output
+    else:
+        return None

lollms/server/endpoints/lollms_infos.py

@@ -1,15 +1,19 @@
 from fastapi import APIRouter
-from lollms_webui import LoLLMSWebUI
+import pkg_resources
+from lollms.server.elf_server import LOLLMSElfServer
+from ascii_colors import ASCIIColors
 
 router = APIRouter()
-lollmsWebUI = LoLLMSWebUI.get_instance()
+lollmsWebUI = LOLLMSElfServer.get_instance()
 
-@router.get("/users")
-def get_users():
-    # Your code here
-    pass
+@router.get("/version")
+def get_version():
+    """
+    Get the version of the lollms package.
 
-@router.post("/users")
-def create_user():
-    # Your code here
-    pass
+    Returns:
+        dict: A dictionary containing the version of the lollms package.
+    """
+    version = pkg_resources.get_distribution('lollms').version
+    ASCIIColors.yellow("Lollms version: " + version)
+    return {"version": version}

lollms/server/server.py

@@ -11,7 +11,7 @@ from fastapi.staticfiles import StaticFiles
 from lollms.app import LollmsApplication
 from lollms.paths import LollmsPaths
 from lollms.main_config import LOLLMSConfig
-from lollms.server.elf import LOLLMSElfServer
+from lollms.server.elf_server import LOLLMSElfServer
 from pathlib import Path
 from ascii_colors import ASCIIColors
 import socketio
@@ -36,7 +36,7 @@ if __name__ == "__main__":
 
     args = parser.parse_args()
     root_path = Path(__file__).parent
-    lollms_paths = LollmsPaths.find_paths(force_local=True, custom_default_cfg_path="configs/config.yaml")
+    lollms_paths = LollmsPaths.find_paths(tool_prefix="elf_",force_local=True, custom_default_cfg_path="configs/config.yaml")
     config = LOLLMSConfig.autoload(lollms_paths)
     if args.host:
         config.host=args.host
@@ -44,6 +44,10 @@ if __name__ == "__main__":
         config.port=args.port
 
     LOLLMSElfServer.build_instance(config=config, lollms_paths=lollms_paths, socketio=sio)
-    from lollms.server.endpoints.lollms_infos import *
+    from lollms.server.endpoints.lollms_infos import router as lollms_infos_router
+    from lollms.server.endpoints.lollms_generator import router as lollms_generator_router
 
+    app.include_router(lollms_infos_router)
+    app.include_router(lollms_generator_router)
+    
     uvicorn.run(app, host=config.host, port=config.port)