Update document_chat.py

document_chat.py

@@ -6,7 +6,10 @@ from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain.chains import ConversationalRetrievalChain
 from langchain.memory import ConversationBufferMemory
 from langchain.llms import HuggingFaceHub
-
+om langchain.chains import ConversationalRetrievalChain
+from langchain.chains.question_answering import load_qa_chain
+from langchain.llms import HuggingFaceHub
+from langchain.memory import ConversationBufferMemory
 # Constants
 CHROMA_DB_PATH = "chroma_db"
 SENTENCE_TRANSFORMER_MODEL = "sentence-transformers/all-MiniLM-L6-v2"
@@ -32,14 +35,22 @@ def ingest_pdf(pdf_path):
     vector_store.add_documents(split_docs)
     vector_store.persist()
 
-def process_query_with_memory(query, chat_history):
+def process_query_with_memory(query, chat_memory):
     """Processes user queries while maintaining conversational memory."""
     retriever = vector_store.as_retriever()
 
-    # Use session memory (should be handled in Streamlit app)
-    qa_chain = ConversationalRetrievalChain(
-        llm=HuggingFaceHub(repo_id=LLM_MODEL, model_kwargs={"max_new_tokens": 500}),
+    # Initialize LLM
+    llm = HuggingFaceHub(repo_id=LLM_MODEL, model_kwargs={"max_new_tokens": 500})
+
+    # Load QA Chain
+    qa_chain = load_qa_chain(llm, chain_type="stuff")
+
+    # Create Conversational Retrieval Chain correctly
+    conversation_chain = ConversationalRetrievalChain.from_llm(
+        llm=llm,
         retriever=retriever,
-        memory=chat_history
+        memory=chat_memory,
+        combine_docs_chain=qa_chain
     )
-    return qa_chain.run({"question": query, "chat_history": chat_history.memory if chat_history else []})
+
+    return conversation_chain.run({"question": query, "chat_history": chat_memory.memory if chat_memory else []})