Update app/main.py

app/main.py

@@ -285,8 +285,6 @@ 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
@@ -346,7 +344,7 @@ async def lifespan(app: FastAPI):
             model_path=model_path,
             temperature=0.3,
             max_tokens=200,
-            n_ctx=1024,
+            n_ctx=1500,
             top_p=0.9,
             top_k=40,
             n_threads=1,
@@ -367,10 +365,10 @@ async def lifespan(app: FastAPI):
 
         retriever = qdrant_vectorstore.as_retriever(search_kwargs={"k": 3})
 
-        combine_docs_chain = RefineDocumentsChain.from_llm(llm=llm)
-        qa_chain = RetrievalQA(combine_documents_chain=combine_docs_chain, retriever=retriever)
+        # 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")
+        qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=retriever, chain_type="stuff")
 
         logger.info("Model and vector store initialized successfully.")
         yield