Update to Use ChatGPT Models (ChatOpenAI)?

[darien-schettler]

Hi @hwchase17 - I noticed in the query_data.py file you are using OpenAI(temperature=0) as your LLM model. I'm just wondering if it's possible to upgrade to using the ChatOpenAI model (so we can use GPT 3.5 or 4 eventually) instead.

I also could be missing something obvious, it happens. Thanks in advance for your time.

[EDIT:]
I implemented this myself, and while it works with minor changes you get the following WARNING:

UserWarning: `ChatVectorDBChain` is deprecated - please use `from langchain.chains import ConversationalRetrievalChain`

However, I tried to implement the ConversationalRetrievalChain and I get the following error:

File "pydantic/main.py", line 341, in pydantic.main.BaseModel.__init__
pydantic.error_wrappers.ValidationError: 1 validation error for ConversationalRetrievalChain retriever instance of BaseRetriever expected (type=type_error.arbitrary_type; expected_arbitrary_type=BaseRetriever)

I assume this is due to ConversationalRetrievalChain using a retriever while ChatVectorDBChain uses a vectorstore directly.

[darien-schettler]

Update. If you want to use ConversationalRetrievalChain with ChatOpenAI you just have to update the inputs so that instead of passing in the vectorstore we just pass in vectorstore.as_retriever().