Source Documents about lanarky HOT 5 CLOSED

Yes of course, let me try to implement this issue

from lanarky.

Hi! do you maybe have an example of the output?

note that I have recently added a new response class for RetrievalQA. I haven't tested if it works with RetrievalQAWithSourcesChain yet.

from fastapi_async_langchain.responses import RetrievalQAStreamingResponse

AFAIK, the sources are added in the end by the LLM so it should be accessible by the on_llm_new_token method of the callback class.

from lanarky.

My output is this, I get sources on last part of message but I would like get page content as is get when return_source_documents is true.

from lanarky.

gotcha! one idea I have is to update the on_chain_end method of the AsyncFastApiStreamingCallback class and stream the source documents using the send attribute. Similar to what the on_llm_new_token method does.

Since this is an interesting feature, would you like to contribute?

from lanarky.

@rogalvil your changes have been published in 0.3.2. Please check if everything is working as expected. If all good, please close this issue. cheers!

from lanarky.