Quick Start
Introduction
PatchPatrol runs AI review as a supported public path: GitLab artifact-first CI. Start by pulling the registry image, wiring the merge request job, and reading the generated artifacts before enabling optional delivery modes.
PatchPatrol is a product by aistack. Visit the PatchPatrol website or contact PatchPatrol for product context.
PatchPatrol has four starting points:
Workspace admin
Prepare registry access, CI variables, and the supported GitLab review job.
Developer user
Run the merge request path and read the artifacts once access is ready.
First output
Open `ai-review.md` first, then use `ai-review.json` when structured detail matters.
Troubleshooting
Recover when setup, artifacts, provider trust gates, or delivery do not match the baseline.
Want to learn more?
New to PatchPatrol? Read What is PatchPatrol?, then return here for the artifact-first rollout steps.
Docker Authentication
Registry access: because GitLab pulls the job image before
scriptruns, configureDOCKER_AUTH_CONFIGas a masked GitLab CI/CD variable, not inside.gitlab-ci.yml.
Create the variable value locally from your registry username and password:
AUTH="$(printf '%s:%s' "$PATCHPATROL_REGISTRY_USERNAME" "$PATCHPATROL_REGISTRY_PASSWORD" | base64 | tr -d '\n')"
printf '{"auths":{"registry.patchpatrol.ai":{"auth":"%s"}}}\n' "$AUTH"Add the printed JSON as the value of a masked DOCKER_AUTH_CONFIG variable in
GitLab project or runner settings. The job below can then pull
registry.patchpatrol.ai/patchpatrol:latest before script starts.
Pipeline Config
# .gitlab-ci.yml (excerpt)
patchpatrol_review:
image: registry.patchpatrol.ai/patchpatrol:latest
rules:
- if: $CI_MERGE_REQUEST_IID
variables:
AI_REVIEW_OUTPUT_DIR: .ai-review
AI_REVIEW_FEEDBACK_MODE: artifact-only
AI_REVIEW_PROVIDER: openai
AI_REVIEW_MODEL: gpt-4o-mini
script:
- ai-review run --mode mr
artifacts:
when: always
paths:
- .ai-review/ai-review.md
- .ai-review/ai-review.jsonReview MR comments
Start with the artifact output created by the merge request job. Open
.ai-review/ai-review.md for the human-readable review and
.ai-review/ai-review.json for structured detail. If you later enable merge
request comments, use Feedback modes after
the artifact-first baseline is working.
# quick validation before the first review run
ai-review test --gitlab-readinessMost clients should use
latestso they receive ongoing review-process improvements. ReplacelatestwithvX.Y.Zonly when you need a specific patch version. KeepAI_REVIEW_FEEDBACK_MODEset toartifact-onlyfor your first rollout. Supported public path: GitLab artifact-first
Continue setup
Operator map
Use these paths when the first artifact-first run is moving from setup into team usage.
For role handoff and optional delivery after the first artifact-first run, use Access and roles and Feedback modes.
If the first run fails, artifacts are missing, or GitLab delivery does not match the expected artifact-first baseline, go straight to Troubleshooting before widening the rollout.
Reference paths
Before rollout
Support contract