llama_index/docs/examples/agent/openai_a...

I think you can just do something like uber_index.insert(document)

Since it's all in memory in pass-by-reference, it should work

@Logan M wdym? We don't pass in an index when we're doing it natively through the openai assistant agent

or at telast, that's what I see here?

ohhh that's the bult-in retrieval assistant (I was looking at the first section lol)

I have no idea in that case

I think you'd just have to re-build the agent with the added data

tbh from what i've seen though, OpenAIs retrieval doesn't work great when you add more files 😅

oh wow really? So you mean that like, openai retrieval with an initial large set of documents works better than an initial smaller set that grows into a larger set?

hmmm rebuilding the agent would lose us the context/memory tho right?

like, I mean it works fine for a small number of documents, much worse with larger. We are working on comaprissons to llama-index 🙂

hmm, true true, unless it's using that threads thing to keep track of chat history? I haven't dove too deep into this specific API myself yet lol

how are the initial comparisons looking like?

worse somewhat?

Ah yeah same here, not sure how upstream memory works will chime back in if I figure it out

yea worse somewhat, and no way to customize or improve it really 😅