[TomCC7]

Custom message projects now share dependency types instead of generating duplicate Header identities. One project config drives dimos build and the Python PEP517/660 backend; built-ins are independently packaged. This authoring follow-up includes generator/distribution exports and package CI prerequisites. Runtime native SDK integration and the coordinator walkthrough belong to the next layer.

Base: #4254. This is layer 4/7, replacing the combined draft #4257 through the user-approved split. Original PR1–3 and the old feat/cdr-runtime-cutover branch/history are preserved. Reassembly uses selected files/hunks and forward merges, not a claim that the original commits cleanly cherry-pick.

Published HEAD: 130f4051b6a209a00696c83441b6b75c14fb2624.

Validation: 32 focused unit tests; compiled Python/C++/Rust cross-package identity and CDR/schema contract; clean isolated source/wheel/sdist/editable/rename installation; offline lock check and hooks passed. The backend wheel was built from candidate sources. GitHub CI is pending.

This remains a proposal with a deliberate old-message API/wire break. CDR replaces encoding; LCM/Zenoh remain transports. No old-recording fallback is promised. No merge or package registry release is requested. External gates remain M20 SDK, host multicast, Foxglove sign-in and human/full-pose visual acceptance. Benchmarks have not been waived or weakened.

Published replacement chain: #4247 → #4252 → #4254 → #4362 → #4363 → #4364 → #4365. Preserved full reference: #4257 at cb982817bde248baa70a6fa69cc9ecc275c43707.

Tutorial follow-up: the checkout workflow is documented in docs/development/messages-in-repository.md; the installed-runtime/custom-project workflow is in docs/development/messages-external-project.md (runtime layer onward); both share message-reference.md. Matching review wheels are required; pip does not resolve uv-only local sources. Clean isolated canonical .msg addition→uv build→wheel install→value/codec checks passed in 112 seconds without installing the DimOS runtime. Existing isolated external source/wheel/sdist/editable acceptance and exact three-language coordinator evidence are retained; embedded C++ includes are now checked correctly. Required toolchains, SDK artifacts and harness acquisition are explicit.

Initial authoring CI exposed a premature root Rust workspace join (lock/Nix source mismatch) and a cold docs build missing Fast CDR. The generated crate remains standalone in authoring and joins the root workspace at runtime cutover; the existing Fast CDR setup now precedes package installation. Locks/assertions/gates were not relaxed. New exact-HEAD CI remains pending.

Authoring namespace test runner now invokes python -m pytest so a clean checkout imports the lightweight message namespace; 11 focused project tests passed. Self-hosted setup now explicitly installs CMake before building Fast CDR, and rosbags==0.11.0 and mcap>=1.2.0 are declared in the shared test group with its locked dependency; eight focused ownership/MCAP cases and hooks passed. Exact updated-head checks remain pending.

Superseded by naming-only replacement #4368. Exact commit 130f4051b6a209a00696c83441b6b75c14fb2624, source tree and parent-relative diff are unchanged.

Verified native seven-layer stack #4372: #4247 → #4366 → #4367 → #4368 → #4369 → #4370 → #4371. Stack: #4247

All branches use cc/ and the ordered PR bases preserve the prior commit ancestry. Old branches/history are retained. Replacement PR checks are pending and prior CI retains its exact SHA/run/ref provenance. No merge, force push or release.

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
5956	1	5955	240

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: 31.03% (Passed 40 times, Failed 18 times)

Stack Traces | 23s run time

voice_bridge = ('http://127.0.0.1:52471/', [])
fake_mic_page = <Page url='http://127.0.0.1:52471/'>

    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)
E       AssertionError: Locator expected to have attribute 'recording'
E       Actual value: arming 
E       Call log:
E         - Expect "to_have_attribute" with timeout 15000ms
E         - waiting for get_by_test_id("chat-audio_in-mic")
E       
E       Aria snapshot:
E       - button "hold to talk": talk

chunks     = []
fake_mic_page = <Page url='http://127.0.0.1:52471/'>
mic        = <Locator frame=<Frame name= url='http://127.0.0.1:52471/'> selector='internal:testid=[data-testid="chat-audio_in-mic"s]'>
url        = 'http://127.0.0.1:52471/'
voice_bridge = ('http://127.0.0.1:52471/', [])

dimos/e2e_tests/test_voice_browser.py:104: 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.}