You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Process sometimes doesn't close after response is finished #4384
I am using a local model with LM Studio, and sometimes (not always) OpenCode waits for further responses, even after it was already finished. You can see this in the interactive as well as the headless mode, where the process also stays open. I then have to abort the operation myself.
To clarify: I do get the actual response, but the process simply doesn't stop after receiving the response.
This issue might be a duplicate of existing issues. Please check:
Bug: opencode run and TUI exits hang on v0.15+ #3213: Bug: opencode run and TUI exits hang on v0.15+ - describes exact same symptoms where processes don't close after completion in both interactive and headless modes
@JimmyBenHur I think this is a preceived hang just because it is waiting for the step-stop, if it is a local llm these things take longer so even though it looks "done" the code is still pending until the llm says it is done
With faster models or providers this shouldn't be an issue
Might be a hang, but I already waited a few minutes and the process wouldn't close.
I'm trying to reproduce it with logs, but right now it's working. Will send them as soon as I can reproduce it.
I had the issue with Qwen3 Coder, that's already a pretty fast model I think.
Activity
github-actions commentedon Nov 16, 2025
This issue might be a duplicate of existing issues. Please check:
opencode runand TUI exits hang on v0.15+ #3213: Bug:opencode runand TUI exits hang on v0.15+ - describes exact same symptoms where processes don't close after completion in both interactive and headless modesFeel free to ignore if none of these address your specific case.
rekram1-node commentedon Nov 16, 2025
@JimmyBenHur I think this is a preceived hang just because it is waiting for the step-stop, if it is a local llm these things take longer so even though it looks "done" the code is still pending until the llm says it is done
With faster models or providers this shouldn't be an issue
rekram1-node commentedon Nov 16, 2025
can u replicate it consistently? Maybe show
opencode run hello --print-logsJimmyBenHur commentedon Nov 16, 2025
Might be a hang, but I already waited a few minutes and the process wouldn't close.
I'm trying to reproduce it with logs, but right now it's working. Will send them as soon as I can reproduce it.
I had the issue with Qwen3 Coder, that's already a pretty fast model I think.
rekram1-node commentedon Nov 16, 2025
Yeah lmk if u get it again