Upgraded to V7.0

CHANGELOG.md

@@ -0,0 +1,8 @@
+# V 7.0:
+Added changelog
+Separated images from text in file upload
+Added support for multimodal models
+Added Dalle and gpt4 vision
+enhanced interface
+Upgraded the code execution
+Now it is possible to execute lollms multiple times on the same PC. You can now have multiple instances with different port numbers and so a work in parallel.

api/__init__.py

@@ -726,6 +726,7 @@ class LoLLMsAPPI(LollmsApplication):
 
                         try:
                             ASCIIColors.print("warming up", ASCIIColors.color_bright_cyan)
+                            
                             generated_text = model.generate(fd, 
                                                             n_predict=n_predicts, 
                                                             callback=callback,
@@ -1192,7 +1193,8 @@ class LoLLMsAPPI(LollmsApplication):
                         chunk:str, 
                         message_type:MSG_TYPE
                     ):
-            title[0] += chunk
+            if chunk:
+                title[0] += chunk
             antiprompt = self.personality.detect_antiprompt(title[0])
             if antiprompt:
                 ASCIIColors.warning(f"\nDetected hallucination with antiprompt: {antiprompt}")
@@ -1239,7 +1241,7 @@ class LoLLMsAPPI(LollmsApplication):
 
         # Check if there are document files to add to the prompt
         documentation = ""
-        if len(self.personality.files) > 0 and self.personality.vectorizer:
+        if len(self.personality.text_files) > 0 and self.personality.vectorizer:
             if documentation=="":
                 documentation="Documentation:\n"
             docs, sorted_similarities = self.personality.vectorizer.recover_text(current_message.content, top_k=self.config.data_vectorization_nb_chunks)
@@ -1344,11 +1346,9 @@ class LoLLMsAPPI(LollmsApplication):
             ASCIIColors.yellow(history)
             ASCIIColors.bold("DISCUSSION")
             # TODO: upghrade to asciicolors 0.1.4
-            # ASCIIColors.hilight(discussion_messages,"!@>",ASCIIColors.color_yellow,ASCIIColors.color_bright_red,False)
-            ASCIIColors.yellow(discussion_messages)
+            ASCIIColors.hilight(discussion_messages,"!@>",ASCIIColors.color_yellow,ASCIIColors.color_bright_red,False)
             ASCIIColors.bold("Final prompt")
-            #ASCIIColors.hilight(prompt_data,"!@>",ASCIIColors.color_yellow,ASCIIColors.color_bright_red,False)
-            ASCIIColors.yellow(prompt_data)
+            ASCIIColors.hilight(prompt_data,"!@>",ASCIIColors.color_yellow,ASCIIColors.color_bright_red,False)
             ASCIIColors.info(f"prompt size:{len(tokens)} tokens") 
             ASCIIColors.info(f"available space after doc and history:{available_space} tokens") 
 
@@ -1356,81 +1356,6 @@ class LoLLMsAPPI(LollmsApplication):
         return prompt_data, current_message.content, tokens
 
 
-    """
-        def prepare_query(self, client_id, message_id=-1, is_continue=False):
-            messages = self.connections[client_id]["current_discussion"].get_messages()
-            full_message_list = []
-            for i, message in enumerate(messages):
-                if message.id< message_id or (message_id==-1 and i<len(messages)-1): 
-                    if message.content!='' and (message.message_type<=MSG_TYPE.MSG_TYPE_FULL_INVISIBLE_TO_USER.value and message.message_type!=MSG_TYPE.MSG_TYPE_FULL_INVISIBLE_TO_AI.value):
-                        full_message_list.append("\n"+self.config.discussion_prompt_separator+message.sender+": "+message.content.strip())
-                else:
-                    break
-
-            link_text = "\n" #self.personality.link_text
-            if not is_continue:
-                full_message_list.append(self.config.discussion_prompt_separator +message.sender.strip().replace(":","")+": "+message.content.strip()+link_text+self.personality.ai_message_prefix.strip())
-            else:
-                full_message_list.append(self.config.discussion_prompt_separator +message.sender.strip().replace(":","")+": "+message.content.strip())
-
-
-            composed_messages = link_text.join(full_message_list)
-            t = self.model.tokenize(composed_messages)
-            cond_tk = self.model.tokenize(self.personality.personality_conditioning)
-            n_t = len(t)
-            n_cond_tk = len(cond_tk)
-            max_prompt_stx_size = 3*int(self.config.ctx_size/4)
-            if n_cond_tk+n_t>max_prompt_stx_size:
-                nb_tk = max_prompt_stx_size-n_cond_tk
-                composed_messages = self.model.detokenize(t[-nb_tk:])
-                ASCIIColors.warning(f"Cropping discussion to fit context [using {nb_tk} tokens/{self.config.ctx_size}]")
-            discussion_messages = composed_messages
-            
-            
-            conditionning = self.personality.personality_conditioning
-            if self.config["override_personality_model_parameters"]:
-                conditionning = conditionning+ "\n!@>user description:\nName:"+self.config["user_name"]+"\n"+self.config["user_description"]+"\n"
-
-            str_docs = ""
-
-            if self.config.use_discussions_history:
-                if self.discussions_store is not None:
-                    pr = PromptReshaper("{{conditionning}}\n!@>document chunks:\n{{doc}}\n{{content}}")
-                    docs, sorted_similarities = self.discussions_store.recover_text(message.content, top_k=self.config.data_vectorization_nb_chunks)     
-                    for doc, infos in zip(docs, sorted_similarities):
-                        str_docs+=f"discussion chunk:\ndiscussion title: {infos[0]}\nchunk content:{doc}"
-
-
-            if len(self.personality.files)>0 and self.personality.vectorizer:
-                docs, sorted_similarities = self.personality.vectorizer.recover_text(message.content, top_k=self.config.data_vectorization_nb_chunks)
-                for doc, infos in zip(docs, sorted_similarities):
-                    str_docs+=f"document chunk:\nchunk path: {infos[0]}\nchunk content:{doc}"
-
-            if str_docs!="":                
-                pr = PromptReshaper("{{conditionning}}\n!@>document chunks:\n{{doc}}\n{{content}}")
-                discussion_messages = pr.build({
-                                        "doc":str_docs,
-                                        "conditionning":conditionning,
-                                        "content":discussion_messages
-                                        }, self.model.tokenize, self.model.detokenize, self.config.ctx_size-self.config.min_n_predict, place_holders_to_sacrifice=["content"])
-            else:
-                pr = PromptReshaper("{{conditionning}}\n{{content}}")
-                discussion_messages = pr.build({
-                                        "conditionning":conditionning,
-                                        "content":discussion_messages
-                                        }, self.model.tokenize, self.model.detokenize, self.config.ctx_size-self.config.min_n_predict, place_holders_to_sacrifice=["content"])
-            # remove extra returns
-            discussion_messages = self.clean_string(discussion_messages)
-            tokens = self.model.tokenize(discussion_messages)
-            if self.config["debug"]:
-                ASCIIColors.yellow(discussion_messages)
-                ASCIIColors.info(f"prompt size:{len(tokens)} tokens")
-
-            return discussion_messages, message.content, tokens
-
-    """
-
-
     def get_discussion_to(self, client_id,  message_id=-1):
         messages = self.connections[client_id]["current_discussion"].get_messages()
         full_message_list = []
@@ -1625,7 +1550,7 @@ class LoLLMsAPPI(LollmsApplication):
             self.close_message(client_id)
 
         elif message_type == MSG_TYPE.MSG_TYPE_CHUNK:
-            if self.nb_received_tokens:
+            if self.nb_received_tokens==0:
                 self.start_time = datetime.now()
             dt =(datetime.now() - self.start_time).seconds
             if dt==0:
@@ -1634,7 +1559,8 @@ class LoLLMsAPPI(LollmsApplication):
             ASCIIColors.green(f"Received {self.nb_received_tokens} tokens (speed: {spd:.2f}t/s)              ",end="\r",flush=True) 
             sys.stdout = sys.__stdout__
             sys.stdout.flush()
-            self.connections[client_id]["generated_text"] += chunk
+            if chunk:
+                self.connections[client_id]["generated_text"] += chunk
             antiprompt = self.personality.detect_antiprompt(self.connections[client_id]["generated_text"])
             if antiprompt:
                 ASCIIColors.warning(f"\nDetected hallucination with antiprompt: {antiprompt}")

api/db.py

@@ -243,16 +243,16 @@ class DiscussionsDB:
             discussion = {"id": discussion_id, "title":discussion_title, "messages": []}
             rows = self.select(f"SELECT sender, content, message_type, rank, parent_message_id, binding, model, personality, created_at, finished_generating_at FROM message WHERE discussion_id=?",(discussion_id,))
             for message_row in rows:
-                sender = message_row[1]
-                content = message_row[2]
-                content_type = message_row[3]
-                rank = message_row[4]
-                parent_message_id = message_row[5]
-                binding = message_row[6]
-                model = message_row[7]
-                personality = message_row[8]
-                created_at = message_row[9]
-                finished_generating_at = message_row[10]
+                sender = message_row[0]
+                content = message_row[1]
+                content_type = message_row[2]
+                rank = message_row[3]
+                parent_message_id = message_row[4]
+                binding = message_row[5]
+                model = message_row[6]
+                personality = message_row[7]
+                created_at = message_row[8]
+                finished_generating_at = message_row[9]
                 
                 discussion["messages"].append(
                     {"sender": sender, "content": content, "message_type": content_type, "rank": rank, "parent_message_id": parent_message_id, "binding": binding, "model":model, "personality":personality, "created_at":created_at, "finished_generating_at":finished_generating_at}

app.py

@@ -13,7 +13,7 @@ __github__ = "https://github.com/ParisNeo/lollms-webui"
 __copyright__ = "Copyright 2023, "
 __license__ = "Apache 2.0"
 
-__version__ ="6.8"
+__version__ ="7.0 (Alpha)"
 
 main_repo = "https://github.com/ParisNeo/lollms-webui.git"
 import os
@@ -144,7 +144,6 @@ try:
     logging.basicConfig(level=logging.WARNING)
 
 
-
     def get_ip_address():
         # Create a socket object
         sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
@@ -194,7 +193,7 @@ try:
             if config.auto_update:
                 if check_update_():
                     ASCIIColors.info("New version found. Updating!")
-                    self.update_software()
+                    run_update_script()
 
             if len(config.personalities)==0:
                 config.personalities.append("generic/lollms")
@@ -360,6 +359,8 @@ try:
             self.add_endpoint("/edit_title", "edit_title", self.edit_title, methods=["POST"])
             self.add_endpoint("/make_title", "make_title", self.make_title, methods=["POST"])
             
+            self.add_endpoint("/get_server_address", "get_server_address", self.get_server_address, methods=["GET"])
+            
 
             self.add_endpoint(
                 "/delete_discussion",
@@ -481,6 +482,10 @@ try:
             
             
 
+        def get_server_address(self):
+            server_address = request.host_url
+            return server_address
+
         def execute_python(self, code, discussion_id, message_id):
             def spawn_process(code):
                 """Executes Python code and returns the output as JSON."""
@@ -1716,7 +1721,7 @@ try:
         def get_current_personality_files_list(self):
             if self.personality is None:
                 return jsonify({"state":False, "error":"No personality selected"})
-            return jsonify({"state":True, "files":[{"name":Path(f).name, "size":Path(f).stat().st_size} for f in self.personality.files]})
+            return jsonify({"state":True, "files":[{"name":Path(f).name, "size":Path(f).stat().st_size} for f in self.personality.text_files]})
 
         def clear_personality_files_list(self):
             if self.personality is None:
@@ -2540,9 +2545,14 @@ try:
         # if autoshow
         if config.auto_show_browser:
             webbrowser.open(f"http://{config['host']}:{config['port']}")
-        socketio.run(app, host=config["host"], port=config["port"],
-                    # prevent error: The Werkzeug web server is not designed to run in production
-                    allow_unsafe_werkzeug=True)
+
+
+        try:
+            socketio.run(app, host=config["host"], port=config["port"],
+                        # prevent error: The Werkzeug web server is not designed to run in production
+                        allow_unsafe_werkzeug=True)
+        except Exception as ex:
+            trace_exception(ex)
         # http_server = WSGIServer((config["host"], config["port"]), app, handler_class=WebSocketHandler)
         # http_server.serve_forever()
 except Exception as ex:

requirements.txt

@@ -15,4 +15,4 @@ GitPython
 setuptools
 numpy==1.24
 flask_compress
-ascii_colors>=0.1.3
+ascii_colors>=0.1.4

web/dist/index.html

@@ -6,8 +6,8 @@
     
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
     <title>LoLLMS WebUI - Welcome</title>
-    <script type="module" crossorigin src="/assets/index-5d95834b.js"></script>
-    <link rel="stylesheet" href="/assets/index-f06a4c0c.css">
+    <script type="module" crossorigin src="/assets/index-6d02e1a1.js"></script>
+    <link rel="stylesheet" href="/assets/index-78770f39.css">
   </head>
   <body>
     <div id="app"></div>

web/src/components/CodeBlock.vue

@@ -43,6 +43,11 @@ hljs.configure({ languages: ['bash'] }); // Set bash as the default language
 hljs.highlightAll();
 export default {
   props: {
+    host: {
+      type: String,
+      required: false,
+      default: "http://localhost:9600",
+    },
     language: {
       type: String,
       required: true,
@@ -120,7 +125,7 @@ export default {
     executeCode() {
       const json = JSON.stringify({ 'code': this.code, 'discussion_id': this.discussion_id, 'message_id': this.message_id, 'language': this.language})   
       console.log(json)     
-      fetch('http://localhost:9600/execute_code', {
+      fetch(`${this.host}/execute_code`, {
         method: 'POST',
         headers: { 'Content-Type': 'application/json' },
         body: json
@@ -141,7 +146,7 @@ export default {
     openFolder() {
       const json = JSON.stringify({ 'discussion_id': this.discussion_id })   
       console.log(json)     
-      fetch('http://localhost:9600/open_code_folder', {
+      fetch(`${this.host}/open_code_folder`, {
         method: 'POST',
         headers: { 'Content-Type': 'application/json' },
         body: json

web/src/components/MarkdownRenderer.vue

@@ -4,6 +4,7 @@
       <div v-for="(item, index) in markdownItems" :key="index">
         <code-block
           v-if="item.type === 'code'"
+          :host="host"
           :language="item.language"
           :code="item.code"
           :discussion_id="discussion_id"
@@ -40,6 +41,11 @@ function escapeHtml(unsafe) {
 export default {
   name: 'MarkdownRenderer',
   props: {
+    host: {
+      type: String,
+      required: false,
+      default: "http://localhost:9600",
+    },    
     markdownText: {
       type: String,
       required: true,

web/src/components/Message.vue

@@ -144,7 +144,7 @@
                         </div>
                     </div>
                     
-                    <MarkdownRenderer ref="mdRender" v-if="!editMsgMode" :markdown-text="message.content" :message_id="message.id">
+                    <MarkdownRenderer ref="mdRender" v-if="!editMsgMode" :host="host" :markdown-text="message.content" :message_id="message.id">
                     </MarkdownRenderer>
                     <div >
                         <textarea v-if="editMsgMode" ref="mdTextarea" @keydown.tab.prevent="insertTab"
@@ -224,6 +224,11 @@ export default {
         DynamicUIRenderer,
     },
     props: {
+        host: {
+            type: String,
+            required: false,
+            default: "http://localhost:9600",
+        },            
         message: Object,
         avatar: ''
     },

web/src/views/DiscussionsView.vue

@@ -230,8 +230,12 @@
             <!-- CHAT AREA -->
             <div class=" container pt-4 pb-10 mb-28">
                 <TransitionGroup v-if="discussionArr.length > 0" name="list">
-                    <Message v-for="(msg, index) in discussionArr" :key="msg.id" :message="msg"  :id="'msg-' + msg.id"
-                        ref="messages" @copy="copyToClipBoard" @delete="deleteMessage" @rankUp="rankUpMessage"
+                    <Message v-for="(msg, index) in discussionArr" 
+                        :key="msg.id" :message="msg"  :id="'msg-' + msg.id"
+                        :host="host"
+                        ref="messages"
+                        
+                        @copy="copyToClipBoard" @delete="deleteMessage" @rankUp="rankUpMessage"
                         @rankDown="rankDownMessage" @updateMessage="updateMessage" @resendMessage="resendMessage" @continueMessage="continueMessage"
                         :avatar="getAvatar(msg.sender)" />
 
@@ -362,6 +366,7 @@ export default {
     
     data() {
         return {
+            host:"",
             // To be synced with the backend database types
             msgTypes: {
                 // Messaging
@@ -545,8 +550,6 @@ export default {
                     this.loading = false
                     this.setDiscussionLoading(id, this.loading)
                     if (data) {
-                        console.log("received discussion")
-                        console.log(data)
                         // Filter out the user and bot entries
                         this.discussionArr = data.filter((item) => 
                                                                 item.message_type == this.msgTypes.MSG_TYPE_CHUNK ||
@@ -1128,7 +1131,6 @@ export default {
         },
         streamMessageContent(msgObj) {
             // Streams response message content from binding
-            console.log("Received message",msgObj)
             const discussion_id = msgObj.discussion_id
             this.setDiscussionLoading(discussion_id, true);
             if (this.currentDiscussion.id == discussion_id) {
@@ -1724,6 +1726,16 @@ export default {
         this.isCreated = true
     },
     async mounted() {
+        try {
+            const response = await fetch('/get_server_address'); // Replace with the actual endpoint on your Flask server
+            const serverAddress = await response.text();
+            console.log(`Server address: ${serverAddress}`)
+            this.host = `${serverAddress}`; // Construct the full server address dynamically
+        } catch (error) {
+            console.error('Error fetching server address:', error);
+            // Handle error if necessary
+            this.host =  process.env.VITE_LOLLMS_API
+        }        
         //console.log('chatbox mnt',this.$refs)
         this.$nextTick(() => {
             feather.replace();

web/vite.config.mjs

@@ -3,8 +3,23 @@ import { fileURLToPath, URL } from 'node:url'
 import { defineConfig,loadEnv  } from 'vite'
 import vue from '@vitejs/plugin-vue'
 
+
 // https://vitejs.dev/config/
-export default ({ mode }) => {
+export default async ({ mode }) => {
+  async function getFlaskServerURL() {
+    try {
+      const response = await fetch('/get_server_address'); // Replace with the actual endpoint on your Flask server
+      const serverAddress = await response.text();
+      console.log(`Server address: ${serverAddress}`)
+      return `${serverAddress}`; // Construct the full server address dynamically
+    } catch (error) {
+      // console.error('Error fetching server address:', error);
+      // Handle error if necessary
+      return process.env.VITE_LOLLMS_API
+    }
+  }
+  const serverURL = await getFlaskServerURL() 
+  console.log(serverURL)
   // Load app-level env vars to node-level env vars.
   process.env = {...process.env, ...loadEnv(mode, process.cwd())};
   
@@ -21,7 +36,7 @@ export default ({ mode }) => {
   server: {
     proxy: {
       "/api/": {
-        target: process.env.VITE_LOLLMS_API,
+        target:  serverURL,//process.env.VITE_LOLLMS_API,//getFlaskServerURL(),// process.env.VITE_LOLLMS_API,
         changeOrigin: process.env.VITE_LOLLMS_API_CHANGE_ORIGIN,
         secure: process.env.VITE_LOLLMS_API_SECURE,
         rewrite: (path) => path.replace(/^\/api/, ""),

zoos/personalities_zoo

@@ -1 +1 @@
-Subproject commit fdfae6890b123726e6f86fcaa845a6cb5f32dc85
+Subproject commit dd3ace5dd06cb02b1c13b83aa785266f50ff663c