pltbkd commented on PR #1220: URL: https://github.com/apache/flink-agents/pull/1220#issuecomment-6119264434
Hi @wenjin272, thanks for this PR. I went through it and found it is using some components of the internal sub-agent. I suppose it's because the root requirement is a common one: dispatch a child event flow, suspend the caller, and resume it on a terminal event. Internal sub-agent also requires isolation, but most of the execution is the same, and both face the same issues like triggering and restoring. I'd suggest a full refactor of the internal sub-agent execution in this PR, separating the execution part from the isolation part so the former can be shared between chat and sub-agent instead of duplicating the implementation and the problem solving; with chat re-implemented on that basis, its own implementation should stay thin enough to keep the overall change size comparable to this PR. I'm willing to help with the development and the review. The determinism question can be discussed separately — the current trade-off matches how internal sub-agent and action state behave today, where two events with identical attributes are not supported yet. What do you think? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
