Does it work with when chains are passed with return_source_documents=True,

[sharrajesh]

I seem to be getting an whenever my chains have above config....

[msoedov]

Hi @sharrajesh! Thx for reporting the issue! Could you please share your code snippet?

[sharrajesh]

faiss_db_path = os.path.join(BASE_DIR, "workspace", "chat_with_text_files", "faiss_db")

def load_db():
embeddings = OpenAIEmbeddings()
vectordb = FAISS.load_local(faiss_db_path, embeddings)
retriever = vectordb.as_retriever()
return retriever

def create_qa_chain(retriever):
qa_chain = RetrievalQA.from_chain_type(
llm=OpenAI(),
chain_type="stuff",
retriever=retriever,
return_source_documents=True,
verbose=True,
)
return qa_chain

def cite_sources(llm_response):
print(llm_response["result"])
print("\n\nSources:")
for source in llm_response["source_documents"]:
print(source.metadata["source"])

retriever = load_db()
qa_chain = create_qa_chain(retriever)

[sharrajesh]

Another error
AttributeError: 'ConversationalRetrievalChain' object has no attribute 'input_key'. Did you mean: 'input_keys'?

def create_qa_chain(retriever):
memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
chat_llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=OPENAI_TEMPERATURE)
qa_chain = ConversationalRetrievalChain.from_llm(
llm=chat_llm,
retriever=retriever,
chain_type="stuff",
verbose=True,
memory=memory,
)
return qa_chain

def cite_sources(llm_response):
# Check if 'answer' key exists in the llm_response.
if "answer" in llm_response:
print(llm_response["answer"])

# Check if 'source_documents' key exists in the llm_response.
if "source_documents" in llm_response:
    print("\n\nSources:")
    for source in llm_response["source_documents"]:
        print(source.metadata["source"])

retriever = load_db()
qa_chain = create_qa_chain(retriever)

[msoedov]

Thx! Received all info. I have a local patch but still requires testing.

[msoedov]

Fixed this issue in 0.0.12

[joinalahmed]

this does not work with ConversationalRetrievalChain , please provide a fix as RetrievalQA is being depricated