The panel's HTTP client does not follow redirects. It asks for
/api/setup/ with a trailing slash, Next answered 308 to normalise it,
and the firmware reported "returned code is not OK. Code - 308" and gave
up. Never having obtained a token, it then called /api/display with an
empty one, got 401, and told the user it could not reach the API.
Not TLS, not the network, not the port — a slash. Two earlier fixes were
aimed at hypotheses the evidence did not support: a certificate chain
the firmware genuinely cannot validate, and a port the shop's network
turned out not to block. Both were reasoned from silence, because
neither Traefik nor a production Next server logs requests by default.
The answer came from a packet capture, and from the device's own words.
/api/log now accepts a report from a panel that cannot authenticate.
Refusing it with a 401 threw away the one diagnostic that mattered: the
firmware was saying exactly what was wrong and we were discarding the
message. Nothing is stored — the rows would reference a device that does
not exist — but it reaches the server log, and the route was already
rate-limited.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012cSY9pVhZmJUKNN7wf1Myd
The e-ink firmware carries a certificate-authority bundle fixed when it
was built, so it cannot validate a chain rooted in an authority created
afterwards. Let's Encrypt's ISRG Root YR was issued in May 2026 and is
not even in an up-to-date Ubuntu CA bundle yet; the kit's firmware
predates it. The handshake fails before a request is ever sent, which is
why neither Traefik nor the application saw anything at all while the
device reported "API connection cannot be established".
Ruled out first, with evidence: TLS 1.2 and the ECDHE-RSA-AES-GCM suites
an ESP32 needs are both offered, and the intermediate is not
cross-signed by an older root, so no alternate path exists in what is
served.
A Traefik router now serves four device paths over :80, ahead of the
entrypoint-wide redirect. The administration stays on TLS. The device
token travels in clear; it is used for nothing else and is revocable
from the settings page, and the image URL is an unguessable content hash.
DEVICE_ALLOW_HTTP existed but was never read — a setting that does
nothing misrepresents what it protects. The device routes now refuse an
unencrypted request unless it is set, so opening this door is a written
decision rather than the silent consequence of a proxy change.
DEPLOY.md records the whole diagnosis, including the commands that
distinguish a TLS failure from an application one, and what to do the
day the firmware learns the new roots.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012cSY9pVhZmJUKNN7wf1Myd
The panel now pairs, fetches its image and files its logs against this
application rather than against the TRMNL cloud.
Four endpoints: /api/setup issues a token on first contact,
/api/display hands back an image and a wake interval, /api/log stores
firmware diagnostics, and /api/device/image/<hash> serves the bytes.
The wake interval is where freshness and battery are traded off. In BYOS
nothing can be pushed: the device sleeps, wakes, asks and sleeps again.
So the interval is short while the shop trades and long overnight, and
it is shortened further whenever a change of state falls inside it —
the door opening in twenty minutes means waking in twenty-one,
whatever the base interval says.
The image filename is the hash of its own bytes. The firmware skips the
redraw when the name is unchanged, which is the whole battery strategy,
and the URL is immutable, unguessable and safe to cache forever. Two
integration tests pin this: unchanged data must yield the same filename
and store one row, changed hours must yield a different one.
MAC addresses are normalised before use. They are a primary key here,
and firmwares are inconsistent about case and separators; without this a
panel could register twice by capitalising itself differently. Header
names are read in both the hyphen and underscore spellings for the same
reason — the TRMNL docs and the Seeed sources disagree, and being
liberal costs nothing while being wrong costs a blank shop window.
Pairing is deliberately made to survive a rendering failure. The token
is issued once and only its digest is kept, so a device stranded by a
failed response would be registered yet hold no credential, and unable
to register again. The welcome image is worth far less than that. This
was found by running the flow, not by reading it.
satori, yoga and harfbuzz are marked external: bundling rewrites the
relative path satori uses to load its WebAssembly, and the renderer dies
on a missing hb.wasm.
The integration tests run against a real Postgres, in CI too. Mocking
Prisma here would only prove the mock works.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012cSY9pVhZmJUKNN7wf1Myd
vliaudatandClaude Opus 5