Update app.py

app.py

@@ -16,6 +16,7 @@ client = Client("on1onmangoes/CNIHUB101324v10", hf_token=HF_TOKEN)
 # Update the conversation history within the function.
 # Return the updated history along with any other required outputs.
 
+@spaces.GPU()
 def stream_chat_with_rag(
     message: str,
     history: list,
@@ -23,123 +24,51 @@ def stream_chat_with_rag(
     system_prompt: str,
     num_retrieved_docs: int = 10,
     num_docs_final: int = 9,
-    temperature: float = 0,
-    max_new_tokens: int = 1024,
-    top_p: float = 1.0,
-    top_k: int = 20,
+    temperature: float = 0, 
+    max_new_tokens: int = 1024, 
+    top_p: float = 1.0, 
+    top_k: int = 20, 
     penalty: float = 1.2,
 ):
     print(f"Message: {message}")
     print(f"History: {history}")
 
+    # OG CODE DELETE
+    # # Add the knowledge Index or VectorStore, RERANKER, 
+    # knowledge_index = vectorstore
+    # reranker = RERANKER
+    
     # Build the conversation prompt including system prompt and history
-    conversation = system_prompt + "\n\n" + f"For Client: {client_name}\n"
-    if history:  # Check if history exists
-        for user_input, assistant_response in history:
-            conversation += f"User: {user_input}\nAssistant: {assistant_response}\n"
-    conversation += f"User: {message}\nAssistant:"  # Add the current message
-
-    # Prepare the data to send to the API
-    api_payload = {
-        "message": conversation,  # Include the history in the message,
-        "history": history,
-        "client_name": client_name,
-        "system_prompt": "",  # Optionally set to empty if included in the message
-        "num_retrieved_docs": num_retrieved_docs,
-        "num_docs_final": num_docs_final,
-        "temperature": temperature,
-        "max_new_tokens": max_new_tokens,
-        "top_p": top_p,
-        "top_k": top_k,
-        "penalty": penalty,
-    }
-
-    try:
-        # Make the API call to get the assistant's reply
-        response = client.predict(
-            api_name="/chat",
-            **api_payload
-        )
-
-        # Extract the assistant's reply
-        if isinstance(response, tuple):
-            answer = response[0]
-        else:
-            answer = response
-
-        # Debugging statements
-        print("The Answer in stream_chat_with_rag:")
-        print(answer)
-
-        # Update the conversation history
-        history.append((message, answer))
-
-    except Exception as e:
-        print(f"An error occurred: {e}")
-        answer = "There was an error retrieving the response."
-
-    # # Return the updated history
-    # return history
-
-
-# def stream_chat_with_rag(
-#     message: str,
-#     history: list,
-#     client_name: str,
-#     system_prompt: str,
-#     num_retrieved_docs: int = 10,
-#     num_docs_final: int = 9,
-#     temperature: float = 0,
-#     max_new_tokens: int = 1024,
-#     top_p: float = 1.0,
-#     top_k: int = 20,
-#     penalty: float = 1.2,
-# ):
-#     print(f"Message: {message}")
-#     print(f"History: {history}")
-
-#     # Build the conversation prompt including system prompt and history
-#     conversation = system_prompt + "\n\n" + f"For Client: {client_name}\n"
-#     for user_input, assistant_response in history:
-#         conversation += f"User: {user_input}\nAssistant: {assistant_response}\n"
-#     conversation += f"User: {message}\nAssistant:"
-
-#     # Prepare the data to send to the API
-#     # Remove 'history' from the payload since the API does not accept it
-#     api_payload = {
-#         "message": conversation,  # Include the history in the message
-#         "client_name": client_name,
-#         "system_prompt": "",  # Optionally set to empty if included in message
-#         "num_retrieved_docs": num_retrieved_docs,
-#         "num_docs_final": num_docs_final,
-#         "temperature": temperature,
-#         "max_new_tokens": max_new_tokens,
-#         "top_p": top_p,
-#         "top_k": top_k,
-#         "penalty": penalty,
-#     }
-
-#     # Make the API call to get the assistant's reply
-#     response = client.predict(
-#         api_name="/chat",
-#         **api_payload
-#     )
-
-#     # Extract the assistant's reply
-#     if isinstance(response, tuple):
-#         answer = response[0]
-#     else:
-#         answer = response
-
-#     # Debugging statements
-#     print("The Answer in stream_chat_with_rag:")
-#     print(answer)
-
-#     # Update the conversation history
-#     history.append((message, answer))
-
-#     # Return the updated history
-#     #return history
+    conversation = system_prompt + "\n\n" + "For Client:" + client_name
+    for user_input, assistant_response in history:
+        conversation += f"User: {user_input}\nAssistant: {assistant_response}\n"
+    conversation += f"User: {message}\nAssistant:"
+
+    # Optionally, if your `answer_with_rag` function or LLM supports context, you can include the conversation
+    # Since you prefer not to modify `answer_with_rag`, we'll proceed with the message as is
+     # OG CODE DELETE
+    # # Call `answer_with_rag` to get the final answer
+    # answer, relevant_docs = answer_with_rag(
+    #     question=message,
+    #     knowledge_index=knowledge_index,
+    #     reranker=reranker,
+    #     num_retrieved_docs=num_retrieved_docs,
+    #     num_docs_final=num_docs_final,
+    #     client_name=client_name,
+    # )
+
+    
+    answer, relevant_docs = client.predict(question=question, api_name="/answer_with_rag")
+    # debug 092624
+    print("The Answer in stream_chat_with_rag:")
+    print(answer)
+
+    print("The relevant_doc:")
+    print(relevant_docs)
+    
+    # Update the conversation history
+    history.append((message, answer))
+    return answer
 
 
 # Function to handle PDF processing API call