Update app.py

app.py

@@ -105,22 +105,6 @@ def generate_message_id():
     return generate_id()
 
 
-def save_conversation(user_id, conversation):
-    """
-    Save conversation history to disk.
-    
-    Args:
-        user_id (str): The unique identifier for the user.
-        conversation (dict): The conversation data.
-        hf_space_path (str): The path to the Hugging Face Space.
-    """
-  
-    # Add the new conversation to the dataset
-    dataset = dataset.add({"user_id": user_id, "conversation": conversation})
-    
-    # Save the updated dataset back to the JSON file
-    dataset.save_to_disk("articko/conversations/conversations.jsonl")
-
 def chat_response(msg_prompt: str) -> dict:
     """
     Generates a response from the model given a prompt.
@@ -154,9 +138,9 @@ def chat_response(msg_prompt: str) -> dict:
     
 
 
-def threaded_chat_response(previous_chat: str, msg_prompt: str) -> dict:
+def thread_response(msg_history: str, msg_prompt: str) -> dict:
     """
-    Generates a response from the model given a prompt.
+    Generates a response from the model using the system prompt written above.
     
     Args:
         msg_prompt (str): The user's message prompt.
@@ -164,14 +148,18 @@ def threaded_chat_response(previous_chat: str, msg_prompt: str) -> dict:
     Returns:
         dict: A dictionary containing the user's message prompt and the model's response.
     """
+    # Validate input parameters
+    if not msg_prompt:
+        raise ValueError("Message prompt cannot be empty.")
+    
     generation_params = {"do_sample": True,"temperature": 0.7,"top_p": 0.95,"top_k": 40,"max_new_tokens": 512,"repetition_penalty": 1.1}
     pipe = pipeline("text-generation",model=model, tokenizer=tokenizer, **generation_params)
-    history_message= f''' incase you are asked about something you talked about the previous days or times or periods,  use this as reference to recall {previous_chat}. Do not point out everything just major keypoints. If nothing is mentioned about the past conversation ingore this. '''
+    message_history= f''' incase you are asked about something you talked about the previous days or times or periods,  use this as reference to recall {msg_history}. Do not point out everything just major keypoints. If nothing is mentioned about the past conversation ingore this. '''
 
     if previous_chat:
         
         prompt_template=f'''
-        <|im_start|>system{system_message}{history_message}<|im_end|>
+        <|im_start|>system{system_message}{message_history}<|im_end|>
         <|im_start|>user {msg_prompt}<|im_end|>
         <|im_start|>assistant
         '''
@@ -181,8 +169,6 @@ def threaded_chat_response(previous_chat: str, msg_prompt: str) -> dict:
         <|im_start|>user {msg_prompt}<|im_end|>
         <|im_start|>assistant
         '''
-
-    
     try:
         
         pipe_output = pipe(prompt_template)[0]['generated_text']
@@ -198,7 +184,7 @@ def threaded_chat_response(previous_chat: str, msg_prompt: str) -> dict:
 
     
 
-def prompt_response(instructions_prompt: str, msg_prompt: str) -> dict:
+def chat_thread_response(ai_persona: str,msg_history: str, msg_prompt: str) -> dict:
     """
     Generates a response based on the provided persona description prompt and user message prompt.
     
@@ -209,31 +195,29 @@ def prompt_response(instructions_prompt: str, msg_prompt: str) -> dict:
     Returns:
         dict: A dictionary containing the user's msg_prompt and the model's response.
     """
-    try:
-        if not instructions_prompt or not msg_prompt:
-            raise ValueError("Instructions prompt template and Message prompt cannot be empty.")
-        
-        # Set generation parameters
-        generation_params = {
-            "do_sample": True,
-            "temperature": 0.7,
-            "top_p": 0.95,
-            "top_k": 40,
-            "max_new_tokens": 512,
-            "repetition_penalty": 1.1
-        }
-        
-        # Create a pipeline for text generation
-        pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, **generation_params)
+
+        # Validate input parameters
+    if not msg_prompt:
+        raise ValueError("Message prompt cannot be empty.")
+    
+    generation_params = {"do_sample": True,"temperature": 0.7,"top_p": 0.95,"top_k": 40,"max_new_tokens": 512,"repetition_penalty": 1.1}
+    pipe = pipeline("text-generation",model=model, tokenizer=tokenizer, **generation_params)
+    message_history= f''' incase you are asked about something you talked about the previous days or times or periods,  use this as reference to recall {msg_history}. Do not point out everything just major keypoints. If nothing is mentioned about the past conversation ingore this. '''
+    
+    if previous_chat:
         
-        # Construct the prompt template
-        prompt_template=f'''<|im_start|>system
-        {instructions_prompt}<|im_end|>
-        <|im_start|>user
-        {msg_prompt}<|im_end|>
+        prompt_template=f'''
+        <|im_start|>system{ai_persona}{message_history}<|im_end|>
+        <|im_start|>user {msg_prompt}<|im_end|>
         <|im_start|>assistant
         '''
-        
+    else:
+        prompt_template=f'''
+        <|im_start|>system{system_message}<|im_end|>
+        <|im_start|>user {msg_prompt}<|im_end|>
+        <|im_start|>assistant
+        '''
+    try:
         # Generate response using the pipeline
         pipe_output = pipe(prompt_template)[0]['generated_text']
       
@@ -268,35 +252,8 @@ async def api_home():
     """
     return HTMLResponse(content=html_content, status_code=200)
 
-@app.post('/chat')
-async def hermes_chat(request: Request):
-    """
-    Starts a new conversation thread with a provided prompt.
-    
-    Args:
-        request (Request): The HTTP request object containing the user prompt.
-    
-    Returns:
-        dict: The response generated by the model.
-    """
-    try:
-
-        data = await request.json()
-        msg_prompt = data.get('msg_prompt')
-
-        if not msg_prompt:
-            raise HTTPException(status_code=400, detail="No prompt provided")
-        response = chat_response(msg_prompt)
-        thread_id = len(conversations) + 1
-        conversations[thread_id] = {'prompt': msg_prompt, 'responses': [response]}
-        return {'thread_id': thread_id, 'response': response}
-    except HTTPException:
-        raise
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=str(e))
-
-@app.post('/chat_thread/{user_id}')
-async def chat_thread(request: Request, user_id: str):
+@app.post('/thread/{user_id}')
+async def thread(request: Request, user_id: str):
     """
     Starts a new conversation thread with a provided prompt for a specific user.
 
@@ -312,13 +269,13 @@ async def chat_thread(request: Request, user_id: str):
 
         data = await request.json()
         msg_prompt = data.get('msg_prompt')
-        previous_chat=data.get('msg_history')
+        msg_history=data.get('msg_history')
 
         if not msg_prompt:
             raise HTTPException(status_code=400, detail="Prompt not provided")
 
         # Generate response
-        response = chat_response(msg_prompt)
+        response = thread_response(msg_history,msg_prompt)
 
         # Generate message ID
         message_id = generate_message_id()
@@ -327,18 +284,19 @@ async def chat_thread(request: Request, user_id: str):
         conversation_thread = {'thread_id': thread_id, 'message_id': message_id,  'message': response}
 
         # Save conversation history to disk
-        history_thread = ({"user_id": user_id, "thread": conversation_thread})
+        thread_history = ({"user_id": user_id, "thread": conversation_thread})
 
-        
         # Return response and thread ID
-        return {'response': history_thread}
+        return {'response': thread_history}
     except HTTPException as e:
         raise e
     except Exception as e:
         raise HTTPException(status_code=500, detail=str(e))
+
         
-@app.post('/prompted_chat')
-async def prompted_chat(request: Request):
+
+@app.post('/chat_thread/{user_id}')
+async def chat_thread(request: Request, user_id: str):
     """
     Starts a new chat thread with a provided user message prompt and persona description of the ai assistant .
     
@@ -349,40 +307,31 @@ async def prompted_chat(request: Request):
         dict: The thread ID and the response generated by the model.
     """
     try:
+        thread_id = generate_thread_id()
+
         data = await request.json()
+        ai_persona= data.get('ai_persona')
         msg_prompt = data.get('msg_prompt')
-        persona_desc = data.get('instructions_prompt')
-
-        if not msg_prompt or not persona_desc:
-            raise HTTPException(status_code=400, detail="Both prompt and person_description are required")
-
-        response = prompt_response(persona_desc, msg_prompt)
+        msg_history=data.get('msg_history')
+       
 
-        thread_id = len(conversations) + 1
-        conversations[thread_id] = {'prompt': msg_prompt, 'responses': [response]}
-        
-        return {'thread_id': thread_id, 'response': response}
-    except HTTPException:
-        raise
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=str(e))
+        if not msg_prompt:
+            raise HTTPException(status_code=400, detail="Prompt not provided")
 
-@app.get('/get_thread/{thread_id}')
-async def get_thread(thread_id: int):
-    """
-    Retrieves the response of a conversation thread by its ID.
-    
-    Args:
-        thread_id (int): The ID of the conversation thread.
-    
-    Returns:
-        dict: The response of the conversation thread.
-    """
-    if thread_id not in conversations:
-        raise HTTPException(status_code=404, detail="Thread not found")
+        # Generate response
+        response = chat_thread_response(ai_persona, msg_history, msg_prompt)
 
-    thread = conversations[thread_id]
-    response = thread['responses'][-1]
+        # Generate message ID
+        message_id = generate_message_id()
 
-    return {'response': response}
+        # Construct conversation entry
+        conversation_thread = {'thread_id': thread_id, 'message_id': message_id,  'message': response}
 
+        # Save conversation history to disk
+        thread_history = ({"user_id": user_id, "thread": conversation_thread})
+        # Return response and thread ID
+        return {'response': history_thread}
+    except HTTPException as e:
+        raise e
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))