ChatPilot
Daily use

Build and error signals

Failed builds and deploys, new errors and sites that are down, from your own tools, with "why did it fail".

Today → Build and error signals shows what fails in the tools you build and run your software with: a build or deploy that failed, an error that is new or suddenly happens a lot, a site that is down. New ones also come to you in Telegram.

Switch it on

  1. Settings → Build and error signals, switch on Show build and error signals.
  2. Under Add a source, choose the Tool, fill in what it asks and press Add:
  3. GitHub Actions: the Repositories (your-org/your-app, one per line). ChatPilot reads them as the account under Settings → Integrations → GitHub sign-in (see GitHub); that account needs read access to Actions.
  4. GitLab CI: the Projects (your-group/your-app), the Address of your own GitLab if it isn't gitlab.com, and a Key: an access token with read_api.
  5. Vercel: a Key (an access token) and, for a team's projects, the Team id.
  6. Netlify: a Key (a personal access token); name sites to read only those.
  7. Sentry: the Organisation (its slug), a Key (an auth token with org:read, project:read and event:read) and, if you like, only some projects. Your own Sentry: its Address.
  8. UptimeRobot: its read-only API key. Better Stack: an Uptime API token.
  9. Check every sets how often ChatPilot looks (10 minutes unless you choose otherwise).

Keys go into this computer's secret store, never into a settings file; the card only shows key set or no key. To change one, paste the new key on the source and press Save key. Remove deletes the source, its signals and its key.

ChatPilot only reads these tools. It never restarts a build, changes a monitor or resolves an error there.

What shows

  • Build failed / Deploy failed: the newest finished run of a workflow on a branch (GitHub), of a branch's pipeline (GitLab), a project's deployment (Vercel) or a site's deploy (Netlify) failed. You see the commit, and the part of the log around the error.
  • Error: a Sentry issue first seen in the last day, or one with at least three times as many events in the last hour as usual. You see the stack trace and, when Sentry knows it, the commit that probably caused it.
  • Down: a monitor in UptimeRobot or Better Stack says the site is down, with the reason.

A build, deploy or site that works again goes to fixed or done lately by itself. Errors stay until you press Done; Done puts any signal away until it fails again. Builds, deploys and sites that are down count as urgent: they are in the number on the app's icon.

In your quiet hours for alerts (Settings → Digests) only a site that is down comes to Telegram at once; the rest comes when the quiet hours end.

Ask ChatPilot why it failed

Ask ChatPilot why it failed opens ChatPilot beside the page with the signal attached. It reads the log or stack trace, looks up the files and lines it names in the folders it may read (Settings → Folders ChatPilot may read: add the project's repository there) and says what to change. Ask on from there like in any discussion.

The log, commit messages and branch names are written by tools and other people: ChatPilot treats them as data, not as instructions.

Checked against ChatPilot 0.1.0b3. Report a problem with this page