Update app/main.py

app/main.py

@@ -285,6 +285,8 @@ from langchain.callbacks.base import BaseCallbackHandler
 from qdrant_client import QdrantClient
 from huggingface_hub import hf_hub_download
 from contextlib import asynccontextmanager
+from langchain.chains.combine_documents import RefineDocumentsChain
+
 
 # === CONFIGURATION === #
 from llama_cpp import Llama
@@ -364,7 +366,11 @@ async def lifespan(app: FastAPI):
         )
 
         retriever = qdrant_vectorstore.as_retriever(search_kwargs={"k": 3})
-        qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=retriever, chain_type="stuff")
+
+        combine_docs_chain = RefineDocumentsChain.from_llm(llm=llm)
+        qa_chain = RetrievalQA(combine_documents_chain=combine_docs_chain, retriever=retriever)
+
+        # qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=retriever, chain_type="stuff")
 
         logger.info("Model and vector store initialized successfully.")
         yield