This polly is closed. Devin Vinson (Polly Sender: @hundreds-twilight-32833) has a polly for you!

For the record, I find the summaries so not useful these days, that I prompt quite a bit something like the attached note: it’s a workaround that makes it repeat the actual LLM response. I find that useful 😂

Looks like it might be worth moving ahead with this change!

I'm not sure which project channel the PR conversation would fit into but this was where the survey came from: https://github.com/OpenHands/OpenHands/pull/13043 . Looks like you've already had a bit of 👀 on it @calm-artist-7206

I think @straight-sundown-84319 has tried to test it. At a cursory look, I think maybe it doesn’t handle the LLM ending with a

finish

event… 🤔 But if Tim thinks it’s fine, it’s fine with me Related, though, it seems that we may be better off if we adjust the agent-server API to return the contents that the agent last had. I filed an issue on that, and started a discussion here: https://openhands-ai.slack.com/archives/C0ACTL5GAQ3/p1775152439739889