[Bug]: Recurring /task timeout after 5 minutes

[shanekunz]

What is broken?

I’m not positive this is a bug in the upstream repo, but my port of the ‘/task’ feature was using ‘session.prompt(...)’ code to spawn the opencode job, and it silently timed out after 5 minutes with no output sent to Telegram.

Steps to reproduce

1. Create a ‘/task’ that often will take longer than 5 minutes. An example is a “Ralph loop” where each agent is trying to complete a big chunk of a long list of tasks.
2. Notice no output is ever delivered.

OS

No response

Additional context (optional)

I believe the fix is to use ‘ promptAsync(...)’. I got it working in uncommitted changes in my fork, but the code is far from solid at this point.

[grinev]

Hmm, I will look. My scheduled task didn't work for 5 minutes yet, so I didn't that

[shanekunz]

@grinev I have basically fixed it locally, but I ended up removing the ...Async() complexity, and I'm not sure if was related to code change or it's specific to an opencode.json setting per machine/repo: https://opencode.ai/docs/config/#models
At the very least, it might be nice if the bot could detect it was a timeout error and link you to the opencode settings docs.