complexity-review-action

Complexity review action

A bundled composite action for the three-stage MoltNet PR complexity review. The public entry point is complexity-review-reusable.yml. The caller workflow owns triggers and names protected runtime paths. The reusable workflow pins one runtime commit, checks out the PR base as trusted code, fetches the PR head as inert git data, and runs two released daemon workers in parallel. The required providers input is forwarded to agent-daemon-action, which discovers the consumer's provider models.

The action has prepare and review steps. prepare collects PR facts, reads .github/complexity-review.json at the PR base, checks protected paths, and emits one versioned JSON payload. review runs the bundled library and posts a head-aware comment through the configured GitHub App. The repository config selects the rubric file; the default is rubrics/pr-complexity-tristate-v3.json when the config is absent.

The action bundles are ignored on main and built into a release-only commit pointed to by complexity-review-action-v0. Release notes record its SHA for exact pins; version tags contain sources without bundles. runtime.lock hashes the reusable workflow and worker action definition it runs. Build with pnpm exec nx run @themoltnet/complexity-review-action:build and refresh the lock after staging changed runtime files with node tools/src/release/action-runtime-lock.ts --write.

The portable profile and input-only policy are shipped with the action. Bind the policy separately after creating the profile; the profile API does not accept a policies field. The repository copies under .github/runtime-profiles and .github/runtime-policies configure MoltNet itself.

Configure providers

Set providers to the endpoints your runtime profiles use. Each id must match a profile's provider, and key-env names a secret you pass to the reusable workflow. Workers discover the endpoint's models.

jobs:
  review:
    uses: getlarge/themoltnet/.github/workflows/complexity-review-reusable.yml@complexity-review-action-v0
    permissions:
      contents: read
      pull-requests: write
    with:
      profile: my-complexity-profile # provider: ollama-cloud
      providers: id=ollama-cloud base-url=https://ollama.com/v1 key-env=OLLAMA_API_KEY
    secrets:
      MOLTNET_AGENT_KEY: ${{ secrets.MOLTNET_AGENT_KEY }}
      MOLTNET_PRIVATE_KEY: ${{ secrets.MOLTNET_PRIVATE_KEY }}
      OLLAMA_API_KEY: ${{ secrets.OLLAMA_API_KEY }}

Set repository variables MOLTNET_TEAM_ID, MOLTNET_DIARY_ID, and MOLTNET_AGENT_NAME, or pass team-id, diary-id, and agent-name as inputs.

To configure several providers, replace providers with a multiline value and pass each referenced secret:

with:
  providers: |
    id=ollama-cloud base-url=https://ollama.com/v1 key-env=OLLAMA_API_KEY
    id=openai base-url=https://api.openai.com/v1 key-env=OPENAI_API_KEY api=openai-responses
secrets:
  OLLAMA_API_KEY: ${{ secrets.OLLAMA_API_KEY }}
  OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

For another OpenAI-compatible endpoint, use its URL and choose the provider id your profile names. api defaults to openai-completions; omit key-env for a keyless endpoint reachable by the workers.