ADRs
0004, Push-to-main deployment
Why Olympus deploys on every push to main rather than via PR review gates
Status: Accepted Date: 2026-02 Stakeholders: Bobby Nannier
Context
Bobby is the sole maintainer and code author across all Olympus repos. PR workflows are an organizational pattern for gating changes, getting another set of eyes before merging. With one author, there are no other eyes.
CI is the gate that matters: lint, typecheck, tests, prod-config verification, image-pin enforcement, secrets scan. CI runs on every push.
Decision
Commit directly to main. No PRs. CI is the gate.
This applies to every Olympus repo: site, athena, hera, platform, sdk, canvas, octl, daedalus.
Consequences
- Fast iteration. A fix lands in seconds, not minutes.
- CI must be green for
mainto be deployable. Red CI means rollback or fix-forward immediately. - No "draft" or staging branches. Use feature flags or unwired routes instead.
- External contributors still PR. This ADR is about the maintainer's flow, not the contribution flow.
Related
- Develop, Release process
- Memory: "Monitor GitHub builds until passing", every push gets watched.
Revisit triggers
- Olympus gains additional maintainers and needs review gates.
- CI itself becomes unreliable.