Tested with the tablet: it picks NCM, not RNDIS, from the Orin's L4T
gadget. Android won't route app traffic over a network without internet
while WiFi or cellular is up, so app-flutter has to request and bind
the Ethernet network. The Orin's USB-C port has no PD, so the tablet is
charged through a PD hub in the car.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The tablet app is the display, so the diagram's hud (local video out)
branch goes; displays are all remote. The car unit now runs no desktop
and no sound server, so the host layer lists ALSA, and mic capture for
Whisper reads it directly.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The car unit has no RTC battery, so before NTP the clock sits near the
last shutdown time; that is the case the two-clock envelope exists for.
The LTE SIM is povo, not an IoT SIM, and its connection now carries
the metered flag the map-sync guard relies on.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
LTE is up over QMI on povo, metered and ranked below Ethernet/WiFi, with
failover tested. CAN interfaces now have stable udev names (can_pcan,
can_onboard) since canN order raced at boot, so references to can1/can0
go. GNSS stays open: gpsd runs but has no fix until an open-sky test,
which is now what blocks the moving map. Kernel-update rebuild steps
are documented for both out-of-tree modules; time sync is pinned to NICT.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two units with different jobs: one builds and runs CI, the other is the
deployment target. Building on the target architecture makes the glibc
constraint structural rather than configured.
The remaining risk is version parity between the two units, not architecture.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013kj58Jugkic783b8ZrtTVZ
Reference, guidelines and plan documents describe what is and what will be.
Retirement stories, resolved-risk narration and against-the-alternative
argument moved out or were cut; reports keep that material.
What survives the cut is the mechanism that can bite again (driving side is a
country-level property) and the condition that would change a decision (a
permanently installed, car-powered panel), both stated generally.
Also removes a duplicated Japan section from a bad edit.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013kj58Jugkic783b8ZrtTVZ
No dedicated display is in scope, so Class A has nothing to drive and app-hud
is cancelled. Records why it will not come back: no connector-powered display
exists (DP_PWR is 3.3V/500mA and deliberately unconnected at the sink, HDMI
gives 55mA, a panel needs 5-15W), the dev kit exposes no DSI, and a data-only
client on one USB cable is simpler anyway.
Video out generated the enclosure, power and cabling complexity this project
set out to solve. Removing it removed the problem.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013kj58Jugkic783b8ZrtTVZ
Replaces `type: default` — one value across every document, read by nothing,
so it discriminated nothing. Types are now reference, guidelines, plan and
report, applied to every markdown file with no exceptions for the hub.
Plans carry status and updated, reports carry date. REPOS.md already noted
that plans go stale and reports do not, but nothing made that visible: a plan
from months ago looked identical to one written today.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013kj58Jugkic783b8ZrtTVZ
Utterances cross as chunks with an explicit terminator, so whole-utterance is
a stream of one chunk and a streaming TTS needs no protocol change. Records
what must not be built — a single-body download-then-play player — since that
forecloses streaming silently.
Also notes the transport is fully exercisable with espeak-ng or a canned WAV,
so audio transport need not wait on GPT-SoVITS.
The provider-behind-an-interface shape has now been picked three times
independently, so it is named as the project idiom rather than re-derived.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013kj58Jugkic783b8ZrtTVZ
The earlier decision assumed the Orin would be the media source. It is not —
the tablet is the music player and that role is staying put, so it holds the
aux cable regardless. Giving the Orin its own aux line would have added a
second permanent cable rather than freeing the display unit.
TTS therefore crosses the boundary as discrete utterances over the reliable
channel rather than a real-time stream, and ducking moves to Android audio
focus, which handles cross-application ducking natively.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013kj58Jugkic783b8ZrtTVZ
arm64 is no longer a risk: verified end to end on mikkeli-orin-nano-2
(L4T R36.4.7). The official image publishes linux/arm64, runs natively, and
the full pipeline works including a served route.
Corrects the Japan guidance. Advising a regional extract was wrong for the
admin build: driving side is a country-level property, and a Shikoku extract
produces no country row and drive_on_right=1, silently giving right-hand
traffic routing. Admins must come from the full extract; routing tiles may
still come from a region.
Decision 14 closed: recordings bounded with rotation, enforced by the writer.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013kj58Jugkic783b8ZrtTVZ
Navigation: self-hosted MapLibre + Valhalla + PMTiles, offline-first, with
Japan specifics (left-hand traffic via valhalla_build_admins, data volume,
CJK labels). Map matching pulled forward as Phase 2 ground truth.
Orin runtime: new document covering the 8 GB shared-memory budget,
capture-first startup tiers, failure isolation, storage and the audio graph.
Data flow: direction belongs to topics inside the Orin; the display boundary
is two simplex channels rather than one duplex socket; cloud links are
outbound-initiated only; services mock themselves rather than having separate
stub implementations.
Naming: repositories are <group>-<thing> with no project prefix, so lexical
order matches semantic grouping. Indexes the five new repos.
Decisions 8-19 recorded; 14 (recording retention) left open.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013kj58Jugkic783b8ZrtTVZ
project-yukart becomes the code-free hub: overview, roadmap and cross-cutting
design only. Each app or app cluster gets its own repo, indexed from here.
- ARCHITECTURE.md: separate services over a message bus, with displays as bus
subscribers so the directly-driven screen, Android client and ESP32 panel
share one contract. Records CAN scope (listen-only + OBD-II as sole writer)
and a decisions table for choices still in flux.
- REPOS.md: repo splitting rule, planned repos, naming, and Gitea template-repo
policy (scaffolding only — no upstream link after instantiation).
- ROADMAP.md: phases 0-5, capture and decode before displays.
Bus choice (MQTT) is marked proposed pending a check against real CAN rates;
implementation language is still open.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
mikkeli
mikkeli