Hi everyone! Just checking in on PR #16913, which fixes chat histories getting mixed when switching conversations. The automated review ran into an error. Should it be retried, or is there anything I should do on my end? Thanks!

Thank you so much for this contribution! Wow oops weird-looking behavior 😅

Thanks, @calm-artist-7206! 😄 I also opened SDK PR #5011 for the subagent isolation issue you reported and reproduced in #4661. Curious whether the implementation matches what you had in mind for that issue. For a quick map of the diff, the main logic is in

registry.py

and

local_conversation.py

. The other production changes connect the conversation-scoped registry to Task/Delegate, remote execution, and resume. About 71% of the added lines are regression tests covering isolation, precedence, and persistence.

I'd love it if @modern-fireman-99755 could take a look at the SDK PR, I appreciate the contribution I’m just idly wondering about the added complexity and where we're storing them for agent-server 🤔 FWIW my agent apparently discovered this during an audit

@calm-artist-7206, I see your point about the complexity. I checked the cases @modern-fireman-99755 pointed out in the review as well. I’ll take another look at the implementation tomorrow and try to work out a simpler approach. will run it by you both before changing the code. also will reach out here if I’m unsure how something should behave. and thanks for the feedback 🙏