Skip to content

TESTING.md

Latest commit

 

History

History
65 lines (48 loc) · 2.45 KB

TESTING.md

File metadata and controls

65 lines (48 loc) · 2.45 KB
Raw

Testing

StreamBuddy has two test suites: one for the Electron app and one for the ML pipeline. Both run on every pull request (see .github/workflows/test.yml).

App (app/)

Uses Vitest and Testing Library.

cd app
npm test            # run everything once
npm run test:watch  # re-run on save
  • test/main/ covers the main process without Electron: clip ownership, the clip folder (list, export, rename, delete), reading the pipeline's progress output, and the local server that finishes Google sign-in.
  • test/renderer/ renders the React screens in a simulated browser with a fake window.electronAPI and a fake Supabase client: login, sign-up, password reset, Google sign-in, session restore, onboarding, the dashboard, settings and export.

ML pipeline (ml/)

Uses pytest.

cd ml
pip install -r requirements-dev.txt
python -m pytest -m "not e2e"   # quick unit tests (no Whisper needed)
python -m pytest -m e2e         # full run on real footage (~1 minute)

The end-to-end tests run pipeline.py the same way the app does and check the finished clip: the stages it reports, 1080x1920 H.264 output with the original audio, the transcript, the webcam at the top, the gameplay in the middle, the blurred background and the burned-in subtitles with the yellow highlight. They also cover 720p recordings, recordings without audio and missing files.

They need ffmpeg, openai-whisper and the Whisper base model (downloaded the first time the pipeline runs), and skip themselves if any of those are missing.

Sample recording

The real-footage checks use the first 12 seconds of a sample recording, which isn't kept in the repository. Use any 1080p gameplay clip with the webcam in the top-left corner and someone talking, and either:

  • save it as ml/examples/test.mp4 (git ignores that folder), or
  • set STREAMBUDDY_SAMPLE_VIDEO to its path.

Without a sample those checks are skipped (as in CI); the checks that build their own test videos still run.

The installer's copy of the pipeline

npm run bundle:pipeline builds a standalone copy of the pipeline for the installer (see the README). To run the same end-to-end checks against that copy instead of pipeline.py, point STREAMBUDDY_PIPELINE_EXE at it:

$env:STREAMBUDDY_PIPELINE_EXE = "..\app\bundle\pipeline\pipeline.exe"
python -m pytest -m e2e
Remove-Item Env:STREAMBUDDY_PIPELINE_EXE