AI Code Review

StationOps can review service code alongside a package build and report findings without blocking the build or automatic deployment. Findings are grouped by severity and can be taken into a fix workspace when an AI-generated change is appropriate.

---

Define Coding Rules

Project administrators can add service-specific guidance under Settings > Code review.

1. Select Add coding rule.
2. Give the rule a clear name.
3. Enter the instruction that generated code and reviews should follow.
4. Select the services to which the rule applies.
5. Save the rule.

Rules may contain up to 20,000 characters. Prefer focused, testable instructions such as architectural boundaries, required validation, or project-specific security constraints.

---

Read Build Findings

For supported service packages, code review starts in parallel with the build. Its status may be pending, in progress, skipped, failed, or succeeded.

A successful review can report:

• Critical findings that should receive immediate attention.
• Warning findings that identify material risks or defects.
• Info findings that suggest smaller improvements.

Each finding can include a title, file and line location, rationale, and recommendation. Review the source context before accepting a recommendation.

---

Propose a Fix

When a review contains fixable findings and the capability is enabled:

1. Select Propose fix from the review results.
2. Open Review proposed fix when generation completes.
3. Inspect every changed file in the fix workspace.
4. Edit the proposed files or ask StationOps to regenerate them with additional instructions.
5. Run the pre-push review to check the current proposal again.
6. Approve or discard the proposal using the workspace actions.

The pre-push review does not push files or create a pull request. It evaluates the proposal currently shown in the workspace so you can make another revision before approval.

---

Related Guide