Enhancements: Better error reporting / status updates

[benagricola]

Okay, I've actually had a chance to use this and compare to kimaki this morning.

So far, pretty good 👍 - although I ran into a couple of issues.

• The bot explicitly discards stdout and stderr from the opencode process in serveManager.ts, making it difficult to debug opencode process failures.
• The bot unconditionally sends a ✅ Done message (and sometimes twice) to Discord when the session becomes idle, even if the command failed. This is an issue for things like endpoint failures (for example if you hit rate limits). Ideally, these would be logged to discord for easy diagnosis.
• The bot does not automatically expand tildes (~) in project paths, leading to ENOENT errors when spawning the shell for worktree creation if the path isn't absolute.
• The Create PR button doesn't do anything, although you can just ask the agent to create the PR itself :)
• Sending a message in /code mode that is very close to the character limit in discord will cause a silent failure (the bot will do nothing). The prompt is first echoed back to the thread with an emoji and a Prompt header, and possibly a message about starting Opencode as well. That message is not truncated to Discord's message character limit (2000) and fails silently, meaning the prompt is never sent to Opencode. This also presents an issue when the agent returns, and the original prompt message is edited to include the reply. If the prompt is long there is very little room for the agent reply to be edited in, so the reply is silently dropped and we just see a 'Done' message.

In general, this feels pretty streamlined already. One thing I liked a lot about kimaki is that it provides regular updates about what the agents are doing (commands they are running, thinking messages etc). Sometimes it can be a bit verbose but it does mean you can interrupt the agent if you see them going off script. It would be nice to have some feedback on what the agents are actually doing (even if it is a toggle so has to be enabled), so it is possible to tell if they are stuck or something.

[RoundTable02]

Hi @benagricola,

Thank you so much for taking the time to write such a comprehensive and detailed issue report—it's incredibly helpful!
I apologize for the delayed response; I've been meaning to get back to this for a while now.

We've since released v1.0.10 (and just pushed v1.1.0 with queue management features), and I wanted to check in to see if you're still experiencing these issues with the latest version.

Quick summary of what's been addressed:

• stdout/stderr logging: This should now be captured and logged properly in serveManager.ts
• Server startup error reporting: Error messages from failed opencode serve processes are now preserved and reported

Items I believe may still need work:

• Tilde (~) expansion in project paths
• The "✅ Done" message behavior on failures
• Character limit handling for long prompts/responses
• Detailed agent activity feedback (thinking messages, commands, etc.)

Could you please confirm if these issues still occur in v1.0.10 or later? If so, it would be tremendously helpful if you could share reproduction steps—for example:

• Which specific command triggers the issue
• Any error messages you see (or don't see)
• Your OS/environment details

I'd love to get these sorted out properly. Thanks again for your patience and for helping improve remote-opencode!