kramlipi CI Repair

Verify-gated CI for GitHub Actions. Pick the symptom — one Action, three experts. Red build → green verify. PR → first-pass review. Diff → run only related tests.

Not a comment farm. Customer-hosted binary. BYO LLM key. We refuse to edit .github/workflows to fake green.

Marketplace: https://github.com/marketplace/actions/kramlipi-ci-repair
Docs: https://kramlipi.github.io/ · Code review CI: https://kramlipi.github.io/code-review-ci/

Pin: @v0.1.7 or @v0 (major). Binary: code-agent-binaries.

---

Symptom → Action (quick pick)

If you feel…	Set expert	What you get
CI is red / build failed	bug-fix	Fix until verify-cmd exits 0; optional draft PR
Reviewers drowning in nits	code-review	Inline first-pass review on every PR
Full suite on a tiny PR	test-intel	Diff-scoped pytest command; skip doc-only MRs
Coverage gate blocks merge	bug-fix	Add tests until --cov-fail-under passes
Same flaky test every week	test-intel + run-tests: true	Flake memory; skip with logged reason

---

Symptom: CI is red / build failed

name: Fix red CI
on:
  workflow_run:
    workflows: [CI]
    types: [completed]
jobs:
  fix:
    if: github.event.workflow_run.conclusion == 'failure'
    runs-on: ubuntu-latest
    permissions:
      contents: write
      pull-requests: write
    steps:
      - uses: actions/checkout@v4
        with:
          ref: ${{ github.event.workflow_run.head_branch }}
          fetch-depth: 0
      - uses: kramlipi/code-agent-action@v0.1.7
        with:
          expert: bug-fix
          verify-cmd: pytest -q
          publish: "true"
        env:
          GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
          CODE_AGENT_LICENSE_KEY: ${{ secrets.CODE_AGENT_LICENSE_KEY }}

Manual / workflow_dispatch:

- uses: kramlipi/code-agent-action@v0.1.7
  with:
    expert: bug-fix
    verify-cmd: go test ./...
    publish: "true"
  env:
    GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}

---

Symptom: PR review backlog

First pass before a senior opens the file. Optional team standards: .code-agent/review-guidelines.md.

name: Review with kramlipi
on:
  pull_request:
    types: [opened, synchronize, reopened, ready_for_review]
jobs:
  review:
    if: github.event.pull_request.draft == false
    runs-on: ubuntu-latest
    permissions:
      contents: read
      pull-requests: write
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0
      - uses: kramlipi/code-agent-action@v0.1.7
        with:
          expert: code-review
          version: code-agent-v0.1.7
        env:
          GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
          GH_TOKEN: ${{ github.token }}
          CODE_AGENT_LICENSE_KEY: ${{ secrets.CODE_AGENT_LICENSE_KEY }}

Working sample: https://github.com/kramlipi/OrderManagementSystem/blob/master/.github/workflows/action.yaml

---

Symptom: Full test suite on every small PR

test-intel maps the PR diff → impacted pytest tests. Doc-only MRs skip the suite. Add .code-agent/coverage.xml on main for sharper selection.

name: Diff-scoped tests
on:
  pull_request:
    types: [opened, synchronize]
jobs:
  test-intel:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0
      - uses: kramlipi/code-agent-action@v0.1.7
        id: intel
        with:
          expert: test-intel
          verify-cmd: pytest -q
          emit-ci: "true"
        env:
          GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
      - name: Run impacted tests only
        run: ${{ steps.intel.outputs.verify_cmd }}

Execute selection + learn flakes:

- uses: kramlipi/code-agent-action@v0.1.7
  with:
    expert: test-intel
    verify-cmd: pytest -q
    run-tests: "true"
    emit-ci: "true"
  env:
    GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}

---

Symptom: Coverage gate blocks merge

- uses: kramlipi/code-agent-action@v0.1.7
  with:
    expert: bug-fix
    verify-cmd: pytest -q --cov=src --cov-fail-under=80
    publish: "true"
  env:
    GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}

---

Inputs

Input	Default	Purpose
expert	bug-fix	bug-fix · code-review · test-intel
verify-cmd	pytest -q	Hard gate for bug-fix / test-intel fallback
publish	false	Open draft PR when true (bug-fix)
base-branch	main	PR / diff base
version	(latest)	Pin code-agent-v0.1.7 from binaries repo
pr	event PR #	code-review / test-intel
guidelines-file	.code-agent/review-guidelines.md	Company MR standards
run-tests	false	test-intel: execute selection + flake memory
emit-ci	false	test-intel: write verify_cmd to GITHUB_OUTPUT
upgrade-url	docs pricing	Free-cap CTA

Secrets

Secret / env	Required	Purpose
GEMINI_API_KEY or OPENAI_API_KEY / ANTHROPIC_API_KEY	Yes	BYO LLM
GH_TOKEN	code-review	${{ github.token }} for inline comments
CODE_AGENT_LICENSE_KEY	After Free CI quota	Team+ for production CI
CODE_AGENT_EMAIL	No	Explicit Free signup identity
CODE_AGENT_INSTALL_ID	No	Default github-actions:<repository_id>
CODE_AGENT_SKIP_ACTIVATE	No	Skip signup

Free CI → paid

First 1000 Free runs need no key (limit remotely updatable). After that, set CODE_AGENT_LICENSE_KEY. One org key may be used on multiple repos.

Run ./code-agent-v0.1.7-linux doctor --provider-test locally to register and get a free run — https://kramlipi.github.io/get-started/

Pricing

• Community (Free) — evaluation / OSS caps
• Team — $49/org/mo
• Business — $199/org/mo

Support: cluevion@gmail.com

Trust

• Subprocess verify-cmd exit 0 is truth (LLM cannot override)
• Workflow YAML edits forbidden
• Binary: kramlipi/code-agent-binaries

License

Apache-2.0 for this Action wrapper. Product terms: https://kramlipi.github.io/