Update document_chat.py

document_chat.py

@@ -28,7 +28,7 @@ def ingest_pdf(pdf_path):
     documents = loader.load()
 
     # Split text into smaller chunks
-    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=50)
     split_docs = text_splitter.split_documents(documents)
 
     # Re-initialize vector store to ensure persistence
@@ -37,7 +37,7 @@ def ingest_pdf(pdf_path):
 
 def process_query_with_memory(query, chat_memory):
     """Processes user queries while maintaining conversational memory."""
-    retriever = vector_store.as_retriever()
+    retriever = vector_store.as_retriever(search_kwargs={"k": 3})  # Limit retrieved chunks
 
     # Initialize LLM
     llm = HuggingFaceHub(repo_id=LLM_MODEL, model_kwargs={"max_new_tokens": 500})
@@ -49,7 +49,7 @@ def process_query_with_memory(query, chat_memory):
         memory=chat_memory
     )
 
-    # Fix: Properly load chat history
-    chat_history = chat_memory.load_memory_variables({}).get("chat_history", [])
+    # Fix: Properly filter chat history to avoid repetition
+    chat_history = list(set(chat_memory.load_memory_variables({}).get("chat_history", [])))
 
-    return conversation_chain.run({"question": query, "chat_history": chat_history})
+    return conversation_chain.run({"question": query, "chat_history": chat_history})