Hey folks 👋 — I just put up a PR I'd love some eyes on: Pluggable in-chat LLM routing → https://github.com/apache/flink-agents/pull/852 The idea: instead of pinning an agent to one chat model, you point it at a little router that picks the model per request and delegates. It's a drop-in CHAT_MODEL, so nothing in the runtime or your agent definition changes. Comes with two strategies out of the box — simple keyword/rule routing and an "ask a cheap judge model" one — plus optional fallback and caching, and you can drop in your own (ML included) without touching the framework. Part two is also in the works: routing records to different agents based on their content. More on that soon. Would really appreciate review on part one. Thanks!

Hi @Purshotam Shah, welcome to the community and thanks for the contribution. The idea sounds interesting. I'd like to take a look. Just to manage the expectation, we are currently busy with finalizing the 0.3 release and preparing for the FFA conference next week, so it might take some time.

Thanks, @Xintong Song! No rush at all — good luck with 0.3 and FFA next week. I'll keep the PR ready, and look forward to your feedback whenever you get to it.