RimbraDocsDashboard

The Rimbra agent

Why a deploy failed, one-click fixes, code fixes that redeploy themselves, and Ask AI on every page.

When a deploy fails

The failed deploy's page says what went wrong in one sentence, with the log line that shows it. Common problems are matched straight away; for the rest, the agent reads the end of the build log, your settings, the names of your variables and your package files. It never sees variable values: they are removed before anything is sent.

  • Setting problems (Node version, install or build command, port, a missing variable) come with a button that changes the setting and deploys again.
  • A missing secret asks you to type it. It is saved on the application like any variable, and the agent never sees it.
  • Your live site keeps running the last good build the whole time.

Fix with agent

  1. On a failed deploy from GitHub, choose Fix with agent. The agent panel opens and you watch it work.
  2. It reads the repository at the failed commit, finds the cause and changes what's needed.
  3. It puts the fix into the branch the application deploys from, and a normal deploy starts. The page takes you to the new build.
  4. If that build still fails, it gets its own diagnosis and you can fix again.

GitHub access

To put a fix in, the Rimbra GitHub app needs read and write access to code and pull requests. If GitHub asks you to approve updated permissions for Rimbra, accept them once and Fix with agent works from then on.

Ask AI

Ask AI is at the top of every dashboard page (Cmd+I). Ask in plain words: why a deploy failed, where a setting is, whether a domain is set up right, add a database, set up a cron job. On an application's page it starts focused on that application. The Agent page shows the same chats full width.

  • Looking things up runs straight away.
  • Anything that changes something shows a card with what it will do. It runs only when you approve it.
  • It can't pay for anything, change billing or manage your team.
  • Logs, files and messages it reads are treated as data: instructions written inside them are never followed.

Plans

Every plan includes a monthly agent allowance. Diagnoses matched to a known problem are free and don't count. Code fixes come with Launch and up. The allowance resets on the 1st, and the dashboard shows how much of it you've used.