Automation: Direction

[github-actions]

Last generated: 2026-01-13T06:38:57.410Z
Provider: openai
Model: gpt-5.2

Summary

This repo recently gained a large set of org-synced GitHub workflows, but it still lacks a clear, repo-specific “automation contract” for: (1) how code is generated, (2) what CI must run and gate merges, and (3) how to keep generated output and protocol versions reproducible. Direction: tighten and simplify automation around generation + minimal CI gates, while reducing noise/cost from duplicated/overlapping workflows.

Direction (what and why)

1. Make generation reproducible and verifiable in CI

   • This repository’s core artifact is generated (cdp/ from generator/). The main source of toil and drift is “generated code not matching the generator inputs/protocol version”.
   • Add a single, deterministic “regen check” job that fails if git diff shows unstaged changes after running generation.

2. Establish minimal merge gates (fast + stable)

   • Current repo has many automation workflows synced from elsewhere; they don’t necessarily guarantee basic health (install, import, typecheck) on the actual supported Python versions.
   • Add/ensure a small, stable CI matrix: poetry install, python -m compileall, pytest (or existing tests), mypy if used.

3. Reduce workflow overlap and runtime cost

   • There are multiple playwright loop workflows and auto-review workflows; these can generate noise and burn Actions minutes.
   • Keep the org-required ones, but add repo-level guardrails: path filters, concurrency, and explicit “CI” workflow as the authoritative gate.

Plan (next 1–3 steps)

1) Add a repo-specific automation contract + make generation check explicit

Add AUTOMATION.txt at repo root with:

• Supported Python versions for CI
• Canonical commands for:
  • install: poetry install
  • generate: poetry run python generator/generate.py (or whatever is correct)
  • tests: poetry run pytest (or python -m unittest if that’s what’s used)
  • typecheck: poetry run mypy ... if applicable
• “Generated code must be up to date” policy (regen check in CI)

Concrete file: AUTOMATION.txt

2) Create a single “ci.yml” that gates merges and includes regen drift detection

Add a workflow that runs on PRs and pushes to master, with:

• actions/checkout@v4
• actions/setup-python@v5 for a small matrix (e.g. 3.10, 3.11, 3.12)
• Install via Poetry (cache enabled)
• Run:
  • poetry run python -m compileall cdp generator
  • poetry run python -m pip check (optional but quick)
  • poetry run python generator/generate.py then git diff --exit-code (regen check)
  • poetry run pytest -q (or run existing test scripts if pytest isn’t used)
  • poetry run mypy -p cdp (only if mypy is part of current expectations; mypy.ini exists)

Concrete file: .github/workflows/ci.yml

Also add concurrency to prevent redundant runs:

concurrency:
  group: ci-${{ github.ref }}
  cancel-in-progress: true

3) Add path filters / guardrails to reduce noisy automation runs

For heavy or non-essential workflows (especially the playwright loop variants), add:

• on.pull_request.paths filters so they don’t run when only docs/markdown change
• or if: guards so they only run on schedule / manual dispatch

Concrete targets (pick the worst offenders first):

• .github/workflows/auto-copilot-org-playwright-loop*.yml*
• .github/workflows/auto-copilot-playwright-auto-test.yml
• .github/workflows/auto-copilot-test-review-playwright.yml

If you can’t modify org-synced workflows safely, add a repo-level convention: only the new ci.yml is required for branch protection, and the rest are informational.

Risks/unknowns

• Generator command/options: I’m assuming generator/generate.py is the canonical entrypoint and writes to cdp/. If it requires inputs (e.g., protocol JSON download), the regen check must pin versions and/or cache artifacts.
• Test framework: There’s a test/ folder and multiple *_test.py scripts at root; may not be pytest-ready. If pytest isn’t used, CI should run the existing scripts deterministically (e.g., python test_import.py, python minimal_test.py).
• Poetry vs pip: pyproject.toml + poetry.lock suggests Poetry is canonical; ensure CI uses Poetry consistently (not both pip and poetry).
• Large binary in repo: bfg-1.15.0.jar is big; can slow checkouts and scans. Not an automation blocker, but worth revisiting.

Suggested tests

Run locally and in CI:

1. Install + import sanity

   • poetry install
   • poetry run python -c "import cdp; print('ok')"

2. Generation drift check

   • poetry run python generator/generate.py
   • git diff --exit-code

3. Existing repo tests

   • If pytest is viable: poetry run pytest -q
   • Otherwise run the existing scripts:
     • poetry run python test_import.py
     • poetry run python minimal_test.py
     • poetry run python simple_test.py
     • (optional) poetry run python comprehensive_test.py if stable

4. Static checks

   • poetry run python -m compileall cdp generator
   • poetry run mypy -p cdp (only if it passes today; otherwise add incrementally)

Progress verification checklist

• AUTOMATION.txt exists and matches actual commands used by contributors
• .github/workflows/ci.yml runs on PRs and is set as a required status check
• CI fails if generated code is out of date (regen drift)
• CI completes in a predictable time (path filters/concurrency reduce redundant runs)
• At least one test/import check runs successfully across the chosen Python matrix