[TomCC7]

Contribution path

• Manipulation SDK preview, #3923.

Problem

• Collection needs to record robot streams with their dataset schema so offline export preserves camera mappings and joint order.

Solution

• Replace the Python collection recorder with a RustRecorder subclass that saves MCAP/SQLite and schema.json from one CollectionProfile.
• Add OpenYAM Quest collection and publish accepted position commands from the base ControlCoordinator.
• Export recordings to LeRobot/HDF5, reconstruct partial command updates, and exclude discarded or incomplete episodes.
• Keep base recorder changes to subclass path resolution and disabled-topic filtering; update message bindings to preserve episode timestamps.
• Migrate xArm/Piper collection; keep G1 on the Python recorder for its shared-memory camera transport.

How to Test

• Run dimos --can-port follower_l run openyam-quest-collection --recorder.recording recordings/session-001 --episodes.task "pick up the cube"; Quest B starts/saves and Y discards.
• 254 stack tests and 9 native integration tests passed, plus mypy and hooks; CI installs libturbojpeg for Python JPEG readback tests.
• Real hardware collection still needs verification.

AI assistance

• OpenAI Codex (GPT-5 and GPT-6) assisted extensively with implementation, tests, documentation, and stack updates; description edited by Codex (agent).

Checklist

• I have read and approved the CLA.

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
6225	1	6224	293

View the full list of 1 ❄️ flaky test(s)

dimos.e2e_tests.test_voice_browser::test_hold_to_talk_ships_a_decodable_recording[firefox]

Flake rate in main: 35.29% (Passed 33 times, Failed 18 times)

Stack Traces | 4.14s run time

voice_bridge = ('http://127.0.0.1:58757/', [AudioChunk(sid='676a352f-a4c3-44b8-8568-8f3b55dd77a5', seq=0, mime='audio/webm;codecs=opu...), AudioChunk(sid='676a352f-a4c3-44b8-8568-8f3b55dd77a5', seq=1, mime='audio/webm;codecs=opus', data=b'', final=True)])
fake_mic_page = <Page url='http://127.0.0.1:58757/'>

    def test_hold_to_talk_ships_a_decodable_recording(
        voice_bridge: tuple[str, list[AudioChunk]], fake_mic_page: Page
    ) -> None:
        url, chunks = voice_bridge
        chunks.clear()  # the module fixture is shared across both engines
        fake_mic_page.goto(url)
        mic = fake_mic_page.get_by_test_id("chat-audio_in-mic")
        # Enabled == transport connected; the manifest already placed the panel.
        expect(mic).to_be_enabled(timeout=120_000)
        expect(mic).to_have_attribute("data-state", "idle")
    
        mic.hover()
        fake_mic_page.mouse.down()
        expect(mic).to_have_attribute("data-state", "recording", timeout=15_000)
        fake_mic_page.wait_for_timeout(1_500)
        fake_mic_page.mouse.up()
        expect(mic).to_have_attribute("data-state", "idle", timeout=30_000)
    
        # The bridge acks each frame only after Out.publish() returned, so by the
        # time the page settled back to idle the port has (about) everything;
        # tolerate the last callback still crossing threads.
        deadline = time.monotonic() + 10.0
        while time.monotonic() < deadline and not any(chunk.final for chunk in chunks):
            time.sleep(0.1)
    
        assert chunks and chunks[-1].final and chunks[-1].data == b""
        assert {chunk.sid for chunk in chunks} == {chunks[0].sid}
        assert [chunk.seq for chunk in chunks] == list(range(len(chunks)))
        assert chunks[0].mime.startswith("audio/")
        audio = b"".join(chunk.data for chunk in chunks)
        assert len(audio) > 1_000  # ~1.5 s of compressed audio, not an empty shell
        # Whatever container this engine produced, the owned ffmpeg decode turns
        # it into the PCM shape the Whisper pipeline assumes.
        event = decode_audio_bytes(audio)
        assert event is not None and event.sample_rate == 16_000
>       assert event.data.ndim == 1 and event.data.shape[0] > 8_000  # > 0.5 s
E       assert (1 == 1 and 7256 > 8000)
E        +  where 1 = array([  0.0013733,   0.0025635,  0.00094604, ...,           0,           0,           0], shape=(7256,), dtype=float32).ndim
E        +    where array([  0.0013733,   0.0025635,  0.00094604, ...,           0,           0,           0], shape=(7256,), dtype=float32) = AudioEvent(shape=(7256,), dtype=float32, sample_rate=16000, channels=1).data

audio      = b'\x1aE\xdf\xa3\x01\x00\x00\x00\x00\x00\x00\x1fB\x86\x81\x01B\xf7\x81\x01B\xf2\x81\x04B\xf3\x81\x08B\x82\x84webmB\x87\...5\xff\xef\x1b\xa1\xcf\xc5\x1d\xeb\x98\xff?\xbf\xe3]/\xf1\'\x8b\x94A\xd2\x1b\x83\x13\x88\xded\xaf\xd0\xc9\xa9\x04L\xe2I'
chunks     = [AudioChunk(sid='676a352f-a4c3-44b8-8568-8f3b55dd77a5', seq=0, mime='audio/webm;codecs=opus', data=b'\x1aE\xdf\xa3\x01...e), AudioChunk(sid='676a352f-a4c3-44b8-8568-8f3b55dd77a5', seq=1, mime='audio/webm;codecs=opus', data=b'', final=True)]
deadline   = 994250.3799415
event      = AudioEvent(shape=(7256,), dtype=float32, sample_rate=16000, channels=1)
fake_mic_page = <Page url='http://127.0.0.1:58757/'>
mic        = <Locator frame=<Frame name= url='http://127.0.0.1:58757/'> selector='internal:testid=[data-testid="chat-audio_in-mic"s]'>
url        = 'http://127.0.0.1:58757/'
voice_bridge = ('http://127.0.0.1:58757/', [AudioChunk(sid='676a352f-a4c3-44b8-8568-8f3b55dd77a5', seq=0, mime='audio/webm;codecs=opu...), AudioChunk(sid='676a352f-a4c3-44b8-8568-8f3b55dd77a5', seq=1, mime='audio/webm;codecs=opus', data=b'', final=True)])

dimos/e2e_tests/test_voice_browser.py:128: AssertionError

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}

[github-actions]

go2 replay realtime

Details

Benchmark suite	Current: 4dc0d8d	Previous: 58acbee	Ratio
first frame wall	9.656 s	9.623 s	1.00
first frame cpu	19.912 s	21.264 s	0.94
run cpu	33.634 s	33.16 s	1.01
run cpu (user)	26.109 s	24.525 s	1.06
run cpu (system)	7.525 s	8.635 s	0.87
peak memory	2741.066 MB	2701.34 MB	1.01
peak threads	415 threads	408 threads	1.02
disk read	33.031 MB	19.965 MB	1.65
disk write	10.895 MB	11.285 MB	0.97
network (transport)	2738.989 MB	2725.519 MB	1.00
network (external rx)	0.063 MB	0.069 MB	0.91
network (external tx)	0.158 MB	0.346 MB	0.46

This comment was automatically generated by workflow using github-action-benchmark.

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'go2 replay realtime'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.08.

Benchmark suite	Current: 4dc0d8d	Previous: 58acbee	Ratio
disk read	33.031 MB	19.965 MB	1.65

This comment was automatically generated by workflow using github-action-benchmark.

[TomCC7]

Replaced by #4343 on a compliant branch at the identical head commit 4dc0d8dccdb941788e5f93a7cf248b3a334ce3b5. The seven layer diffs and downstream PR IDs were verified unchanged in native stack #4345: #4343 → #4344 → #3921 → #3931 → #3942 → #4088 → #4156.

Closing this PR only for the user-approved branch naming migration. This branch and all review comments remain available; unresolved findings and existing main/CI blockers are not marked resolved. No behavior changes were included in this migration.