The fleet has ~700 devices and no way to see them: the device console manages one
machine from its own keyboard, and the question "which devices are installed, and
which wave is each in" had no surface at all. This adds it as the first screen.
Python standard library only -- http.server, html.escape, string.Template. No
framework and no third-party import, because the fleet installs offline from a
committed package pool that contains no web framework, so fastapi/uvicorn/flask
would have made the panel uninstallable on the devices it manages. That was
decided with the user over the FastAPI they first picked, and a test asserts it:
the module imports only stdlib plus its own siblings.
What it shows: the dark header with the target and the privilege level; a summary
of counts by state with one loud figure; the device list with a state badge per
row, monospace ids and hostnames so columns line up, non-ready devices sorted
first, and a filter bar; the selected device's detail with its four-channel
verification, since READY is refused until every channel is verified (docs/27).
Nothing is colour-only: every badge carries its Turkish state word.
Data has no live source yet, so a missing inventory file falls back to synthetic
records that the page labels as demo in two places -- the panel never presents
invented devices as real ones.
Two defects found by reviewing a real render, both fixed here with a test that
fails when the bug returns (proven by mutation: each assertion goes red against
the reintroduced bug, green after):
- --state-provisioned resolved to var(--info), and --info is --accent, so a
KURULDU badge and the selected row were the same blue. The state now resolves
to its own colour, and the test resolves var() aliases before comparing --
comparing the raw declarations does not detect this, which is how the first
version of the test passed against the bug.
- the wave cell printed the bare code and then a label that already contains it,
so one device read "P2 P2 — PILOT-2". Each wave is named once, now.
Also fixes two entries in docs/27 that had become untrue: it claimed a clone
carries the packages and the installer downloads nothing, which stopped being the
case when the pool moved to Git LFS and bf-bootstrap began cloning without
smudge. The real flow -- light clone, payload from a local mirror or a targeted
`git lfs pull` -- is documented instead.
Tests: tests/test_web_static.py, 112 tests, collected by tests/run-all.sh
(12/12 PASS overall).
The device console is a default-grey tkinter window with no hierarchy: no
grouping, no state colour, no type scale, the privilege warning buried in a
monospace paragraph, "Çık" duplicated at two sizes, and an output area that
looks editable. Every measurement is in the delivered work; the point here is
that the fix is a system, not a repaint.
Adds PRODUCT.md (product truth: who uses this, from where, under what
constraints), docs/DESIGN.md (the direction contract and the shared status
vocabulary) and web/static/tokens.css (the values both surfaces read).
The vocabulary is the load-bearing part. Four severities -- ok / warn / danger /
info -- plus one action accent, and one Turkish word per device stage
(KURULDU / KAYITLI / HAZIR / HATA / YOK) used identically on the device and in
the central panel. A fifth hue is a bug, and colour is never the only carrier:
every badge ships its word, so the fleet list stays readable in greyscale.
The ground is light on purpose. The console runs in workshops and dealer service
areas in daylight; a near-black ops console loses its own body text there and its
state colours collapse into equal-brightness smudges.
No webfont: the fleet installs offline from a committed pool, so a font CDN is a
broken page on every device.
