Update app.py

app.py

@@ -21,115 +21,61 @@ def stream_chat_with_rag(
     history: list,
     client_name: str,
     system_prompt: str,
-    num_retrieved_docs: int,
-    num_docs_final: int,
-    temperature: float,
-    max_new_tokens: int,
-    top_p: float,
-    top_k: int,
-    penalty: float,
+    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,
 ):
-    # Use the parameters provided by the UI
+    print(f"Message: {message}")
+    print(f"History: {history}")
+
+    # Build the conversation prompt including system prompt and history
+    conversation = system_prompt + "\n\n" + "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
+    api_payload = {
+        "message": message,
+        "history": history,
+        "client_name": client_name,
+        "system_prompt": system_prompt,
+        "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(
-        message=message,
-        client_name=client_name,
-        system_prompt=system_prompt,
-        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,
-        api_name="/chat"
+        api_name="/chat",
+        **api_payload
     )
 
-    # Debugging statements
-    print(f"API Response Type: {type(response)}")
-    print(f"API Response: {response}")
+    # Assuming the API returns the assistant's reply
+    # If the API returns a tuple (answer, relevant_docs), extract the answer
+    if isinstance(response, tuple):
+        answer = response[0]
+    else:
+        answer = response  # If it's already the assistant's reply
 
-    # Ensure response is a string
-    if isinstance(response, list):
-        # If response is a list, extract the assistant's reply
-        response = response[0] if response else ""
+    # Debugging statements
+    print("The Answer in stream_chat_with_rag:")
+    print(answer)
 
     # Update the conversation history
-    history = history + [(message, response)]
+    history.append((message, answer))
 
     # Return the updated history
     return history
 
-
-
-
-# # Function to handle chat API call
-# # Function to handle chat API call
-# def stream_chat_with_rag(
-#     message: str,
-#     history: list,
-#     client_name: str,
-#     system_prompt: str,
-#     num_retrieved_docs: int,
-#     num_docs_final: int,
-#     temperature: float,
-#     max_new_tokens: int,
-#     top_p: float,
-#     top_k: int,
-#     penalty: float,
-# ):
-#     # Use the parameters provided by the UI
-#     response = client.predict(
-#         message=message,
-#         client_name=client_name,
-#         system_prompt=system_prompt,
-#         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,
-#         api_name="/chat"
-#     )
-    
-#     # Update the conversation history
-#     history = history + [(message, response)]
-    
-#     # Return the assistant's reply and the updated history
-#     return "", history
-
-
-# # Function to handle chat API call
-# def stream_chat_with_rag(
-#     message: str,
-#     history: list,
-#     client_name: str,
-#     system_prompt: str,
-#     num_retrieved_docs: int,
-#     num_docs_final: int,
-#     temperature: float,
-#     max_new_tokens: int,
-#     top_p: float,
-#     top_k: int,
-#     penalty: float,
-# ):
-#     # Use the parameters provided by the UI
-#     response = client.predict(
-#         message=message,
-#         client_name=client_name,
-#         system_prompt=system_prompt,
-#         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,
-#         api_name="/chat"
-#     )
-#     # Return the assistant's reply
-#     return response
-
 # Function to handle PDF processing API call
 def process_pdf(pdf_file):
     return client.predict(