Updated code

app.py

@@ -30,7 +30,7 @@ from langchain.agents.agent_toolkits import create_conversational_retrieval_agen
 from langchain.utilities import SerpAPIWrapper
 
 from utils import build_embedding_model, build_llm
-from utils import load_ensemble_retriver, load_text_chunks, load_vectorstore, load_conversational_retrievel_chain
+from utils import load_ensemble_retriver,load_vectorstore, load_conversational_retrievel_chain
 
 load_dotenv()
 # Getting current timestamp to keep track of historical conversations
@@ -51,11 +51,11 @@ if "embeddings" not in st.session_state:
 if "vector_db" not in st.session_state:
     st.session_state["vector_db"] = load_vectorstore(persist_directory=persist_directory, embeddings=st.session_state["embeddings"])
  
-if "text_chunks" not in st.session_state:
-    st.session_state["text_chunks"] = load_text_chunks(text_chunks_pkl_dir=all_docs_pkl_directory)
+# if "text_chunks" not in st.session_state:
+#     st.session_state["text_chunks"] = load_text_chunks(text_chunks_pkl_dir=all_docs_pkl_directory)
 
 if "ensemble_retriver" not in st.session_state:
-    st.session_state["ensemble_retriver"] = load_ensemble_retriver(text_chunks=st.session_state["text_chunks"], embeddings=st.session_state["embeddings"], chroma_vectorstore=st.session_state["vector_db"] )
+    st.session_state["ensemble_retriver"] = load_ensemble_retriver(embeddings=st.session_state["embeddings"], chroma_vectorstore=st.session_state["vector_db"] )
 
 if "conversation_chain" not in st.session_state:
     st.session_state["conversation_chain"] = load_conversational_retrievel_chain(retriever=st.session_state["ensemble_retriver"], llm=st.session_state["llm"])    

utils.py

@@ -48,7 +48,7 @@ def build_llm():
     Loading OpenAI model
     '''
     # llm= OpenAI(temperature=0.2)
-    llm= ChatOpenAI(temperature = 0, max_tokens=256)
+    llm= ChatOpenAI(temperature = 0)
     return llm
 
 def build_embedding_model():
@@ -253,15 +253,15 @@ def load_text_chunks(text_chunks_pkl_dir):
             pickle.dump(all_texts, file)
         print("Text chunks are created and cached")
         
-def load_ensemble_retriver(text_chunks, embeddings, chroma_vectorstore):
+def load_ensemble_retriver(embeddings, chroma_vectorstore):
     """Load ensemble retiriever with BM25 and Chroma as individual retrievers"""
-    bm25_retriever = BM25Retriever.from_documents(text_chunks)
-    bm25_retriever.k = 2
-    chroma_retriever = chroma_vectorstore.as_retriever(search_kwargs={"k": 3})  
-    ensemble_retriever = EnsembleRetriever(retrievers=[bm25_retriever, chroma_retriever], weights=[0.3, 0.7])
+    # bm25_retriever = BM25Retriever.from_documents(text_chunks)
+    # bm25_retriever.k = 2
+    chroma_retriever = chroma_vectorstore.as_retriever(search_kwargs={"k": 10})  
+    # ensemble_retriever = EnsembleRetriever(retrievers=[bm25_retriever, chroma_retriever], weights=[0.3, 0.7])
     logging.basicConfig()
     logging.getLogger('langchain.retrievers.multi_query').setLevel(logging.INFO)
-    retriever_from_llm = MultiQueryRetriever.from_llm(retriever=ensemble_retriever,
+    retriever_from_llm = MultiQueryRetriever.from_llm(retriever=chroma_retriever,
                                                               llm=ChatOpenAI(temperature=0))
     return retriever_from_llm
 
@@ -322,6 +322,6 @@ def load_conversational_retrievel_chain(retriever, llm):
         chain_type="stuff",
         retriever=retriever,
         return_source_documents=True,
-        chain_type_kwargs={"prompt": prompt, "memory": memory},
+        chain_type_kwargs={"memory": memory},
     )
     return qa