hey y'all, sorry if this is a simple

Or is this a perfect LangChain -> LlamaIndex opportunity?

You should be able to create the router query engine, wrap it in a query engine tool, and give it to an agent 👀

soon as i wrap it in a query tool and pass it into a ContextChatEngine as a retriever the answers that come back are how i should use to the tools to get the answer i want, not the answer itself :/

Ohhh it won't work in context chat engine. That one uses a retriever only, not a full query engine.

You would need to use a router retriever for that one

An agent would work fine with a query engine though

ah, ic

what im looking for is the memory bit that the chat engine offers

such that i can put a chat ui on top of it

thank you Logan!

An agent also has memory 🙂

Which tbh with an agent, you might not even need the router (just a query engine tool for each underlying query engine)

just looking at the classes for agents, am i limited to openai with agents?

Nope, you can use a react agent for everything else non-openai

this here ya?

.... beautiful

You got it.

But your milage may vary. Openai is quite good compared to open source for agent-like applications

from what i was reading the zephyr models are my best bet?

Nailed it again, was just going to suggest that 👍

I think we even have a notebook for setting it up

Agents at the bottom, LLM config at the top

lovely.... i was working form this very notebook last week, missed the agent stuff entirely 😂

thank you again Logan