Upload 2 files

retriever/chat_manager.py

@@ -1,43 +1,57 @@
+from datetime import datetime
 import logging
 from typing import List
 from globals import app_config
 
-def chat_response(query: str, selected_docs: List[str], history: str) -> str:
+def chat_response(query: str, selected_docs: List[str], history: List[dict]) -> List[dict]:
     """
     Generate a chat response based on the user's query and selected documents.
 
     Args:
         query (str): The user's query.
         selected_docs (List[str]): List of selected document filenames from the dropdown.
-        history (str): The chat history.
-        model_name (str): The name of the LLM model to use for generation.
+        history (List[dict]): The chat history as a list of {'role': str, 'content': str} dictionaries.
 
     Returns:
-        str: Updated chat history with the new response.
+        List[dict]: Updated chat history with the new response in 'messages' format.
     """
+    timestamp = datetime.now().strftime("%H:%M:%S")
+
+    # Handle empty query
     if not query:
-        return history + "\n" + "Response: Please enter a query." if history else "Response: Please enter a query."
+        return history + [{"role": "assistant", "content": "Please enter a query."}]
 
+    # Handle no selected documents
     if not selected_docs:
-        return history + "\n" + "LLM: Please select at least one document." if history else "Response: Please select at least one document."
+        return history + [{"role": "assistant", "content": "Please select at least one document."}]
 
     # Retrieve the top 5 chunks based on the query and selected documents
     top_k_results = app_config.doc_manager.retrieve_top_k(query, selected_docs, k=5)
     
     if not top_k_results:
-        return history + "\n" + f"User: {query}\nLLM: No relevant information found in the selected documents." if history else f"User: {query}\nLLM: No relevant information found in the selected documents."
+        return history + [
+            {"role": "user", "content": f"{query}"},
+            {"role": "assistant", "content": "No relevant information found in the selected documents."}
+        ]
 
     # Send the top K results to the LLM to generate a response
     try:
         llm_response, source_docs = app_config.gen_llm.generate_response(query, top_k_results)
     except Exception as e:
-        return history + "\n" + f"User: {query}\nLLM: Error generating response: {str(e)}" if history else f"User: {query}\nLLM: Error generating response: {str(e)}"
-
-    # Format the response for the chat history
-    response = f"{llm_response}\n"
-    '''for i, doc in enumerate(source_docs, 1):
-        doc_id = doc.metadata.get('doc_id', 'Unknown')
-        filename = next((name for name, d_id in app_config.doc_manager.document_ids.items() if d_id == doc_id), 'Unknown')
-        response += f"{i}. {filename}: {doc.page_content[:100]}...\n"'''
-
-    return history + "\n" + f"User: {query}\nResponse: {response}" if history else f"User: {query}\nResponse: {response}"
+        return history + [
+            {"role": "user", "content": f"{query}"},
+            {"role": "assistant", "content": f"Error generating response: {str(e)}"}
+        ]
+
+    # Format the response (uncomment and adapt if you want to include source docs)
+    response = llm_response
+    # for i, doc in enumerate(source_docs, 1):
+    #     doc_id = doc.metadata.get('doc_id', 'Unknown')
+    #     filename = next((name for name, d_id in app_config.doc_manager.document_ids.items() if d_id == doc_id), 'Unknown')
+    #     response += f"\n{i}. {filename}: {doc.page_content[:100]}..."
+
+    # Return updated history with new user query and LLM response
+    return history + [
+        {"role": "user", "content": f"{query}"},
+        {"role": "assistant", "content": response}
+    ]

retriever/document_manager.py

@@ -104,10 +104,11 @@ class DocumentManager:
         top_k_results = all_results[:k]
 
         # Log the list of retrieved documents
-        logging.info("Retrieved top K documents:")
+        logging.info(f"Result from search :{all_results} ")
+        logging.info(f"Retrieved top {k} documents:")
         for i, result in enumerate(top_k_results, 1):
             doc_id = result['metadata'].get('doc_id', 'Unknown')
             filename = next((name for name, d_id in self.document_ids.items() if d_id == doc_id), 'Unknown')
-            logging.info(f"{i}. Filename: {filename}, Doc ID: {doc_id}, Score: {result['score']:.4f}, Text: {result['text'][:100]}...")
+            logging.info(f"{i}. Filename: {filename}, Doc ID: {doc_id}, Score: {result['score']:.4f}, Text: {result['text'][:200]}...")
 
         return top_k_results