Orbi

• 

  Orbi Cloud setup: add the ai-ready label to dispatch the first Issue.

Inspiration

Orbi keeps software delivery in GitHub. An Issue states the goal and acceptance criteria; the pull request records the implementation and review. GitHub Issues hold the task state, so running the engine does not require a separate task database or queue.

What it does

Label an Issue ai-ready to start implementation in an isolated Git worktree. Orbi runs tests and opens a pull request. A separate AI session reviews the change and fixes findings before Orbi checks the merge gates. Repository policies can require maintainer action before merging. Repository maintainers request releases separately; Orbi then runs the release gates and publishes the tag and release notes.

How it is built

The engine is written in Python. It uses the Pi coding agent for implementation and review, Git worktrees to separate tasks, and the GitHub CLI to work with Issues and pull requests. A scheduler starts the runner on a timer. The engine is open source under AGPL-3.0 and can be self-hosted. Orbi Cloud runs it in managed sandboxes.

Engineering choices

If a pull request changes after review, the previous verdict no longer describes its latest commit. Orbi binds the merge decision to the reviewed commit. Recoverable failures continue on the same pull request; failures that need a human decision leave the Issue marked ai-blocked.

PR #1023 records a review finding about rollback after a release tag had already been pushed. The maintainer confirmed the finding, and the PR was revised before merge. The discussion and subsequent commits show how the implementation changed.

Website: https://orbi.build/?ref=dir-devpost

Source: https://github.com/orbi-build/orbi

Built With

• git
• github
• python