Streaming

I would say make sure streaming works with the plain index first, then try to figure out how that can connect to langchain lol

Streaming in llama index can be a little weird, since internally it can make many llm calls, changing its answer along the way

I think only the last answer is streamed

Id love to be able to contribute to llama index and i suppose that starts with understanding langchain properly 🫡

Sounds like a plan! 😎

this may be why sometimes it streamed the json string of the output with sources (like i wanted) and othertimes it just streamed the initial 'thoughts' and then stopped? that would make sense

but nice one thankyou