Workspace stuck in RUNNING status after task completes (opencode and pi agents)

[haruki-06]

Environment

Docker (Docker host running on Proxmox)
OpenAlice: Latest version
AI provider: Google Gemini Developer API
Model: gemini-3.1-flash-lite
Symptom
When I ask Alice a question in a new workspace, the workspace status gets stuck on RUNNING for tens of minutes and never changes. No response appears in the chat.
The issue occurs with both the opencode and pi agents. Recreating the workspace does not resolve the issue.
What I've found so far

Gemini API: The AI Studio usage dashboard shows that all API requests are succeeding without errors.
Agent log: The opencode log (opencode.log) shows around 9 processing steps and ends with an "exiting loop" line. This suggests that the agent itself has finished processing the task normally.
OpenAlice log: The main OpenAlice log (docker logs) stops immediately after transcript.session.captured. Nothing else is logged after that.
Process status: The opencode process remains alive even after the "exiting loop" line is printed (confirmed via /proc), so the process does not actually terminate.
Reproducibility: The issue reproduces consistently across container restarts and newly created workspaces, so it does not appear to be an isolated or workspace-specific issue.
I also took a quick look at transcript-watcher.ts, but wasn't able to trace what's supposed to happen after transcript.session.captured.
Note: I'm not very technical and used an AI assistant (Claude) to help investigate the logs and put this report together. Just wanted to flag this in case it's a real bug — happy to gather more info or run any commands if that would help.

[luokerenx4]

I think it's some daemon progress issues. I will check it quickly.

[luokerenx4]

I can't re-construct this case. You can make an issue so that I can catch up with the further fix.