[simonvanlierde]

This moves the H264 lores encoder methods out of the core CameraBackend protocol and into StreamingCameraBackend. Before this, every backend had to implement them, even a stills-only sensor with no preview stream. A backend without them crashed the preview pipeline with an AttributeError.

• CameraBackend declared start_lores_encoder and stop_lores_encoder. They now live on StreamingCameraBackend. StreamingNotSupportedError moved next to the protocols, and the manager imports it from there.
• The preview pipeline called the hooks on any CameraBackend. It now narrows to StreamingCameraBackend and raises StreamingNotSupportedError for a backend without them.
• The preview sleeper and thermal governor passed any backend on to the pipeline. They now skip a backend without the hooks, so a core-only camera does not log a warning every tick.
• The worker and pipeline test mocks were plain MagicMocks, which fail a runtime protocol check. They now use a StreamingCameraBackend spec. I also updated the backend paragraph in ARCHITECTURE.md.

No test runs Picamera2Backend on a Pi, so the hardware path is unchecked. Picamera2Backend already subclasses StreamingCameraBackend and keeps both methods unchanged.

Tests:

• just check: exit 0

  Required test coverage of 80.0% reached. Total coverage: 94.87%
  ======================== 840 passed, 1 skipped in 8.03s ========================
  

  ruff and ty check both reported "All checks passed!".
• ! sed -n '/^class CameraBackend/,/^class StreamingCameraBackend/p' app/camera/services/backend.py | grep -q 'lores_encoder': exit 0, no output.

I ran the two new tests on the code before the change and both failed. test_start_raises_streaming_not_supported failed with AttributeError: '_StillOnlyBackend' object has no attribute 'start_lores_encoder'. test_tick_is_noop_for_backend_without_lores_encoder failed with Expected 'start' to not have been called. Called 1 times.

---

[codspeed]

Merging this PR will not alter performance

1 warning

⚠️ 2 benchmarks are dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

✅ 10 untouched benchmarks

---

_{Comparing move-lores-encoder-streaming (c88559d) with main (da58389)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!