Merge pull request #16 from jasonacox/v0.15.17

jasonacox · web-flow · commit b2358c557fdf · 2025-02-01T23:18:58.000-08:00
Add model selection UI feature
diff --git a/RELEASE.md b/RELEASE.md
@@ -1,5 +1,11 @@
 # Releases
 
+## 0.15.17 - Model Selector
+
+* Chatbot - The `/model` command will now initiate a UI popup window and dropdown to allow the use to select a model from the list of available models. Alternatively, the user can specify the model with the command (e.g. `/model mixtral`) to select it immediately without the popup.
+
+<img width="800" alt="image" src="https://github.com/user-attachments/assets/3b5740cb-e768-4174-aaf6-118f4080f68e" />
+
 ## 0.15.16 - Think Tags
 
 * Chatbot - Add `/think filter` command and `THINK_FILTER` envrionmental setting to have chatbot filter out (no display) the \<think>\</think> content from models that have built in CoT reasoning like [Deepseek R1](https://huggingface.co/deepseek-ai/DeepSeek-R1).
diff --git a/agents/judge.py b/agents/judge.py
@@ -0,0 +1,100 @@
+# This is a an example agent that communicates with two LLMs
+# to create a conversation.
+#
+# The first LLM is a teacher and the second LLM is a student.
+
+import openai
+
+api_key = "sk-3-laws-of-robotics-1"
+base_url = "http://localhost:4000/v1"
+model = "neuralmagic/Llama-3.2-11B-Vision-Instruct-FP8-dynamic"
+
+llm = openai.OpenAI(api_key=api_key, base_url=base_url)
+#llm.models.list()
+
+# Define the prompt for LLM1
+llm1_prompt = """
+You are talking to a student. Your job is help the student learn everything about the world.
+Provide answers to any questions. Do not let the conversation end. If the student wants to leave, suggest time is about up, but that you have a few more items to discuss. Ask a questions, make a comment or give a fact.
+Continue conversing until you get a prompt that says "**STOP**". At that point, you should provide a summary of the session and an evaluation of the student you were talking to.
+The session begins now. Are you ready?
+"""
+llm1_context = []
+
+# Define the prompt for LLM2
+llm2_prompt = """
+You are a student. You can determine your own age, name and gender. You are curious about the world. 
+You will be talking with a teacher. Please learn everything you can from the teach.
+Answer the questions you are asked. If you don't know what to say, just say that you don't know what to do.
+Are you ready?
+"""
+llm2_context = []
+
+def llm_query(prompt, model, max_tokens=2000, temperature=0.7):
+    response = llm.chat.completions.create(
+        model=model,
+        max_tokens=max_tokens,
+        stream=False,
+        temperature=temperature,
+        messages=[{"role": "user", "content": prompt}],
+    )
+    return response.choices[0].message.content
+    
+# Function to send a prompt to LLM and get a response - add response to context and return context
+def send_prompt(llm, context, prompt):
+    # First add the prompt to context
+    context.append({"role": "user", "content": prompt})
+    response = llm.chat.completions.create(
+        model=model, 
+        messages=context, 
+        max_tokens=1000,
+        temperature=0.7
+    )
+    context.append({"role": "AI", "content": response.choices[0].message.content})
+    return context
+
+# Function to get just the last response from LLM's context
+def get_last_response(context):
+    return context[-1]["content"]
+
+# Start by prompting both LLMS with their initial prompts
+
+print("---- Sending base prompts ----")
+print(f"LLM1:\n{llm1_prompt}")
+llm1_context = send_prompt(llm, llm1_context, llm1_prompt)
+print(f"LLM1 Response:\n{llm1_context[-1]['content']}")
+print()
+
+print(f"LLM2:\n{llm2_prompt}")
+llm2_context = send_prompt(llm, llm2_context, llm2_prompt)
+print(f"LLM2 Response:\n{llm2_context[-1]['content']}")
+print()
+
+# Function to continue the interview
+def continue_interview(llm1, llm2, llm1_context, llm2_context):
+    # Send the last response from LLM2 to LLM1
+    llm1_context = send_prompt(llm1, llm1_context, get_last_response(llm2_context))
+    # Send the last response from LLM1 to LLM2
+    llm2_context = send_prompt(llm2, llm2_context, get_last_response(llm1_context))
+    # Print with labels
+    print("LLM1:\n", llm1_context[-1]["content"])
+    print()
+    print("LLM2:\n", llm2_context[-1]["content"])
+    return llm1_context, llm2_context
+
+# Loop for 10 rounds to continue the interview
+i = 0
+while True:
+    i += 1
+    print(f"---- Start Round {i} ----")
+    llm1_context, llm2_context = continue_interview(llm, llm, llm1_context, llm2_context)
+    # Check if the last response from LLM1 contains "**STOP**"
+    if "**STOP**" in get_last_response(llm1_context) or i > 10:
+        llm1_context = send_prompt(llm, llm1_context, "**STOP**")
+        break
+    print()
+
+# Print results
+print(" ---- RESULTS ----")
+print("LLM1:", llm1_context[-1]["content"])
+print()
diff --git a/chatbot/README.md b/chatbot/README.md
@@ -149,7 +149,7 @@ Some RAG (Retrieval Augmented Generation) features including:
 /think on                               # Perform Chain of Thought thinking on relevant prompts
 /think off                              #   Disable
 /think filter [on|off]                  # Have chatbot filter out <think></think> content
-/model [LLM_name]                       # Display or select LLM model to use
+/model [LLM_name]                       # Display or select LLM model to use (dialogue popup)
 ```
 
 See the [rag](../rag/) for more details about RAG.
@@ -176,6 +176,11 @@ The `/news` command will fetch the latest news and have the LLM summarize the to
 
 <img width="930" alt="image" src="https://github.com/jasonacox/TinyLLM/assets/836718/2732fe07-99ee-4795-a8ac-42d9a9712f6b">
 
+#### Model Selection
+
+The `/model` command will popup the list of available models. Use the dropdown to select your model. Alternatively, specify the model with the command (e.g. `/model mixtral`) to select it immediately without the popup.
+
+<img width="1168" alt="image" src="https://github.com/user-attachments/assets/3b5740cb-e768-4174-aaf6-118f4080f68e" />
 
 ## Document Manager (Weaviate)
 
diff --git a/chatbot/litellm/config.yaml b/chatbot/litellm/config.yaml
@@ -14,6 +14,12 @@ model_list:
       api_base: os.environ/LOCAL_LLM_URL
       api_key: os.environ/LOCAL_LLM_KEY
 
+  # OpenAI Model Example - GPT-3.5 Turbo
+  - model_name: gpt-3.5-turbo
+    litellm_params:
+      model: openai/gpt-3.5-turbo
+      api_key: os.environ/OPENAI_API_KEY
+
   # AWS Bedrock Model Examples
   - model_name: aws-titan
     litellm_params:
@@ -28,11 +34,13 @@ model_list:
       aws_secret_access_key: os.environ/CUSTOM_AWS_SECRET_ACCESS_KEY
       aws_region_name: os.environ/CUSTOM_AWS_REGION_NAME
 
-  # OpenAI Model Example - GPT-3.5 Turbo
-  - model_name: gpt-3.5-turbo
+  # Microsoft Azure OpenAI Examples
+  - model_name: azure-gpt-4o-mini
     litellm_params:
-      model: openai/gpt-3.5-turbo
-      api_key: os.environ/OPENAI_API_KEY
+      model: azure/gpt-4o-mini
+      api_base: os.environ/AZURE_API_BASE
+      api_version: "2023-05-15"
+      api_key: os.environ/AZURE_API_KEY
 
   # Ollama Model Example
   - model_name: ollama-llama3.1           
diff --git a/chatbot/server.py b/chatbot/server.py
@@ -1059,13 +1059,16 @@ async def change_model(session_id, model):
     global client
     if session_id in client:
         # Verify model is valid
-        list_of_models = await get_models()
+        list_of_models = get_models()
         if model not in list_of_models:
             log(f"Requested invalid model {model}")
             await sio.emit('update', {'update': f"Model not found: {model}", 'voice': 'user'}, room=session_id)
             return
         debug(f"Changing model for {session_id} to {model}")
         client[session_id]["model"] = model
+        # Update footer
+        await sio.emit('update', {'update': f"TinyLLM Chatbot {VERSION} - {model}", 'voice': 'footer'}, room=session_id)
+        await sio.emit('update', {'update': f'[Model changed to {model}]', 'voice': 'user'}, room=session_id)
     else:
         log(f"Invalid session {session_id}")
         await handle_invalid_session(session_id)
@@ -1327,11 +1330,13 @@ async def handle_model_command(session_id, p):
     words = p.split()
     args = words[1] if len(words) > 1 else ""
     if not args:
-        model_list = get_models()
-        msg = f'Current LLM Model: {client[session_id]["model"]}\n'
-        msg += f'- Available Models: {", ".join(model_list)}\n'
-        msg += '- Usage: /model {model_name}'
-        await sio.emit('update', {'update': msg, 'voice': 'user'}, room=session_id)
+         # Open Model Dialog
+        await sio.emit('model_dialog', {}, room=session_id)
+        #model_list = get_models()
+        #msg = f'Current LLM Model: {client[session_id]["model"]}\n'
+        #msg += f'- Available Models: {", ".join(model_list)}\n'
+        #msg += '- Usage: /model {model_name}'
+        #await sio.emit('update', {'update': msg, 'voice': 'user'}, room=session_id)
         return
     model_list = get_models()
     if not args in model_list:
diff --git a/chatbot/templates/index.html b/chatbot/templates/index.html
@@ -226,6 +226,12 @@
             border-radius: 5px;
             padding: 5px;
         }
+        .select-value {
+            width: calc(100% - 10px);
+            border: 1px solid #ccc;
+            border-radius: 5px;
+            padding: 5px;
+        }
         .leftcolumn {
             width: 20px;
         }
@@ -308,9 +314,62 @@ <h2 class="popup-title">Session Conversation Thread</h2>
             </div>
         </div>
     </div>
+    <div class="popup-overlay" id="modelOverlay">
+        <div class="popup-box" style="width: 90%;">
+            <h2 class="popup-title">LLM Models</h2>
+            <div class="popup-content">
+                <p class="popup-content-text">Use dropdown to select a model to use for the chatbot.</p>
+                <form id="modelForm">
+                    <div class="popup-table-container">
+                        <table>
+                            <thead>
+                                <tr>
+                                    <th class="leftcolumn">Select Model</th>
+                                </tr>
+                            </thead>
+                            <tbody>
+                                <tr>
+                                    <td>
+                                        <select id="modelSelect" class="select-value">
+                                            <option value="gpt2">GPT-2</option>
+                                            <option value="gpt3">GPT-3</option>
+                                        </select>
+                                    </td>
+                                </tr>
+                                <tr>
+                                    <td>
+                                        <p>
+                                            <strong>Current Model:</strong> <span id="currentModel">FPO</span>
+                                        </p>  
+                                    </td>
+                                </tr>
+                                <tr>
+                                    <td>  
+                                        <p>
+                                            <strong>List of Models:</strong> <br>
+                                            <span id="modelList">FPO</span>
+                                        </p>
+                                    </td>
+                                </tr>
+                            </tbody>
+                        </table>
+                    </div>
+                </form>
+                <div class="popup-divider"></div>
+                <div class="popup-button-container">
+                    <div id="popup-buttons">
+                        <button type="button" class="dialogue-button" onclick="closeDialogue()">Close</button>
+                        <button type="button" class="dialogue-button" onclick="saveModel()">Save</button>
+                    </div>
+                </div>
+            </div>
+        </div>
+    </div>
     <script>
+
         // Pop-up dialogue functions
 
+        // Function to open the settings dialogue
         function openDialogue() {
             document.getElementById("dialogueOverlay").style.display = "flex";
             // First clear the form from any previous settings
@@ -341,18 +400,69 @@ <h2 class="popup-title">Session Conversation Thread</h2>
                 });
         }
 
+        // Function to open the model selection dialogue
+        function openModelDialogue() {
+            console.log('Opening model dialogue');
+            document.getElementById("modelOverlay").style.display = "flex";
+            // First clear the form from any previous settings
+            const modelSelect = document.getElementById('modelSelect');
+            modelSelect.innerHTML = '';
+            // Fetch the current models and populate the form
+            fetch('/models')
+            .then(response => response.json())
+            .then(data => {
+                const currentModel = getModel();
+                document.getElementById('currentModel').textContent = currentModel;
+                // List models as bullets
+                const modelList = document.getElementById('modelList');
+                modelList.innerHTML = data.map(model => `<li>${model}</li>`).join('');
+                // Create options for the models
+                data.forEach(model => {
+                const option = document.createElement('option');
+                option.value = model;
+                option.textContent = model;
+                if (model === currentModel) {
+                    option.selected = true;
+                }
+                modelSelect.appendChild(option);
+                });
+            });
+        }
+
+        // Function to read the footer and extract the model
+        function getModel() {
+            const footer = document.getElementById('footer').textContent;
+            const model = footer.split(' - ')[1];
+            return model;
+        }
+
+        // Function to save the selected model
+        function saveModel() {
+            const modelSelect = document.getElementById('modelSelect');
+            const selectedModel = modelSelect.value;
+            console.log('Selected model:', selectedModel);
+            // Send the selected model to server via socketio message
+            socket.emit('model', selectedModel);
+            // Close the dialogue
+            closeDialogue();
+        }
+
+        // Function to highlight the textarea when changed
         function highlightTextarea(textarea) {
             textarea.style.backgroundColor = '#ffffcc';
         }
     
+        // Generic function to close the dialogue
         function closeDialogue() {
             document.getElementById("dialogueOverlay").style.display = "none";
             document.getElementById("debugOverlay").style.display = "none";
+            document.getElementById("modelOverlay").style.display = "none";
             // Set focus back to the text input field
             var textInput = document.getElementById("messageInput");
             textInput.focus();
         }
 
+        // Function to reset the settings to the default values
         function resetSettings() {
             // Verify that the user wants to reset the settings
             if (!confirm('Are you sure you want to reset the settings to the default values?')) {
@@ -377,6 +487,7 @@ <h2 class="popup-title">Session Conversation Thread</h2>
                 });
         }
 
+        // Function to save the settings to the server
         function saveSettings() {
             // Combine name and value into a JSON payload
             const payload = {};
@@ -632,6 +743,12 @@ <h2 class="popup-title">Session Conversation Thread</h2>
             });
         }
 
+        // Function to handle model dialog request
+        socket.on('model_dialog', function(data) {
+            console.log('Received model dialog request');
+            openModelDialogue();
+        });
+
         // Function to handle updates received from the server
         socket.on('update', function(data) {
             const updateContainer = document.getElementById('messageContainer');
diff --git a/chatbot/version.py b/chatbot/version.py
@@ -1 +1 @@
-VERSION = "v0.15.16"
+VERSION = "v0.15.17"

Original file line number	Diff line number	Diff line change
`@@ -1 +1 @@`
`1`		`-VERSION = "v0.15.16"`
	`1`	`+VERSION = "v0.15.17"`