Agent and multishot queries

[Lawouach]

Hello,

I was wondering what would be the best approach to implement an agent streaming a multishot query?

When I'm using a query pipeline, I basically take the output of a previous query response and inject into the next query as part of the payload. That works well but I'd like to stream the responses as they arrive and it seems Agents are the place for streaming.

So, what would be the right approach here?

[timonv]

Hey, good point, I haven't implemented it (yet) for the query pipeline. Streaming relies on complete_stream from the ChatCompletion trait, only agents use it at the moment. Would indeed be super nice to have, I'll create an issue.

[timonv]

Created in #826