stable-diffusion-webui/CONTRIBUTING.md

Contributing to Serena

Serena is a governed refactor program for AUTOMATIC1111/stable-diffusion-webui. This guide helps you run verification locally and understand CI behavior.

---

Quickstart

1. Install dependencies: pip install -r requirements-test.txt and runtime deps per requirements_versions.txt
2. Create stub repositories: python scripts/dev/create_stub_repos.py
3. Run tests: see Local verification below

---

Local verification

To verify the project builds and tests pass without a real model:

python launch.py --skip-prepare-environment --skip-torch-cuda-test --exit
pytest test

For full CI parity (including server startup and API tests), start the test server in one terminal, then run pytest in another. See .github/workflows/run_tests.yaml for the exact commands CI uses.

---

CI parity

CI uses deterministic stub repositories and fake inference to avoid external clones and real model loading. To reproduce CI locally:

1. Run python scripts/dev/create_stub_repos.py before any launch or test
2. Use --skip-prepare-environment when launching
3. CI sets CI=true (GitHub Actions default); txt2img/img2img return a deterministic 1×1 PNG in CI instead of invoking the model

---

Stub repositories

CI uses stub repositories to satisfy import paths without cloning large external repos (e.g. ldm, sgm). See scripts/dev/create_stub_repos.py for implementation. The script creates minimal placeholder modules so the application can start and API contract tests can run.

---

Development workflow

1. Create a branch from main
2. Make changes; run ruff . and pytest test locally
3. Open a PR; CI runs linter, tests, and coverage
4. Do not push directly to main; merge via PR after CI passes

For milestone-specific workflow, see docs/serena.md and docs/milestones/.