The three mutation gates matched source text that has since moved or been
reflowed, so 37 of their 108 reverts reported STALE and could no longer
prove anything:
- qa-floorhold-mutations (2): the neighbour-floor code moved from
src/data/groundFloor.js into src/services/groundFloor.js (ac977bd).
- qa-flyroute-mutations (8): src/cameraVerbs.js was reflowed by the
formatter (2aca7ae); the anchors now match the formatted lines.
- qa-firstrun-mutations (27): the launcher markup, styles and startup moved
to src/ui/templates/welcome.html, src/ui/styles/first-run.css and
src/app/startupChrome.js; the rail default to panelPositionControls.js and
the radio ESC handling to radioBindings.js (a67bc60 and follow-ups). The
tool-schema pin now edits a tool description, since the context-mode enum
it used to edit no longer lives in tools.js.
Every revert still reintroduces the same defect. Results: floor-hold 22/22,
fly_route 19/19, first-run 67/67 caught, with all source files restored.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The Path 3 section assumed readers already knew how a bind-mounted container
persists state (14traits, #208). It now lists what the app saves and where,
with the NAS and reverse-proxy detail in the collapsed block that was already
there, so the open part stays about the same length.
It also fixes two commands that did not do what they said in bash, confirmed
in bash and dash by PoBruno: `UID=1026 GID=100 docker compose up` prints
"UID: readonly variable" and still runs as the caller's UID, and
`export UID GID` exports an empty GID because bash does not define one. Both
are now `env UID=... GID=...`, which `docker compose config` resolves to the
requested ids.
Suggested-by: 14traits
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Two findings from PoBruno's Docker Engine on Linux run (#208), and one from
re-running the suite in the container after rebasing:
- The launcher pinned its own allowedHosts, including '.local'. Vite reads a
leading dot as a suffix match, so `Host: anything.local` reached the page and
the data proxies. Main has since moved the policy into build/allowedHosts.js,
which drops suffix entries on purpose and applies before the app's routes, so
the launcher's override is deleted rather than patched. As a side effect
GEV_ALLOWED_HOSTS now also works from `.env`, as it does under `npm run dev`.
- A checkout the container cannot write - typically a clone made as root on a
NAS or Proxmox host - made Save Keys answer 500 and never created .gev-cache,
with nothing logged. The launcher now warns at start, naming the owner, the
ids it runs as, and the two fixes.
- On Docker Desktop the bind-mounted repo always shows an owner other than the
container user, so git refuses it as "dubious ownership" and the tests that
shell out to git fail (src/data/bundledJson.test.mjs, new on main). The image
now marks /app, and only /app, as a safe directory.
Verified in a container on this base: `anything.local` 403 on the page and on
/api routes, localhost 200, GEV_ALLOWED_HOSTS honoured from the environment and
from .env, the warning on a root-owned /app run as 1000:1000, and
bundledJson.test.mjs 6/6.
Reported-by: PoBruno
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Docker delivers the host's own browser to the container from the container
network's gateway, never from 127.0.0.1, so the loopback-only admission gate
answered every POWER UP request with 403 and no key could be saved. The README
and CHANGELOG claimed otherwise; both are corrected here.
- keySetupCore: GEV_KEY_SETUP_TRUSTED_PEERS, an exact-address list (no
wildcards, prefixes or hostnames) admitted alongside loopback. Unset trusts
nothing, and the sharing, Host, Origin, Content-Type and proxy-header checks
all still apply to a listed address. Twelve assertions pin each refusal.
- scripts/docker-start.mjs: the container entry point, shaped after
scripts/pinokio-start.mjs. It reads the default gateway from /proc/net/route,
trusts exactly that address, and starts Vite. No gateway (--network none)
trusts nothing. The pure helpers are unit-tested and a contract test pins the
Dockerfile CMD and the loopback-only publish.
- Dockerfile: install git. `npm test` builds fixtures with `git init`, so the
suite failed 71 tests inside the image and the README's claim that it runs in
the container was wrong.
- SECURITY.md and .env.example document the setting and why it is safe only
behind a 127.0.0.1 publish; README gains the Linux/NAS id note and .dockerignore
drops macOS AppleDouble sidecars, both reported on the PR.
Verified in Docker on a clean build: saving a key from the panel writes the
bind-mounted .env (0600 in the container), it survives `compose down`/`up`, and
REMOVE comments it back out. A foreign Origin, a foreign Host, a proxy header
and a sibling container on the same network are each refused. `npm run doctor`,
`npm test` (5067 passed, 1 skipped, 0 failed) and `npm run build` are green
inside the container.
Reported-by: @jklaze
Co-Authored-By: Bogdan Petrut <[email protected]>
Co-Authored-By: Claude Opus 5 <[email protected]>
Vite's dev server is the whole app here — it serves the client and runs every
API proxy as middleware — so one container running `npm run dev` covers it, and
HOST/PORT are already env-driven.
Three details are deliberate:
- The port publishes to 127.0.0.1. Docker's usual "4173:4173" binds every host
interface, which would silently make the README's LAN warning the default for
a server that brokers configured provider keys.
- The repo is bind-mounted, so keys saved from Provider Settings, .gev-cache
and .gev-logs persist on the host exactly as they do under `npm run dev`.
node_modules stays in the image so a host install cannot shadow the
Linux-built native deps.
- The container runs as the host user, since everything it writes lands in
that bind mount. Vite's dep optimizer still needs to write inside the image's
node_modules, hence the two loosened directories.
Puppeteer's Chromium download is skipped: it is only needed by `npm run
test:track`, which the README points at the host.
Co-Authored-By: Claude Opus 5 <[email protected]>
Review of #966: with a Google Maps key but no Mapillary token, switching
Street Level on read KEY REQUIRED / "Needs MAPILLARY_CLIENT_TOKEN" with
nothing about Street View, which stayed off (billing), so the layer looked
broken.
- While the switched-on providers lack their keys, the layer reports the
switched-off ones that would draw with a key already set
(keyAlternatives: Street View on Google 3D with a Google key). The
panel's KEY REQUIRED tooltip, its coverage line and the layer row's
tooltip then read "Needs MAPILLARY_CLIENT_TOKEN — add it in Provider
Settings, or light STREET VIEW to use your Google Maps key"
(keySetupWithAlternatives). Off Google 3D the plain KEY REQUIRED stays.
Street View stays off by default.
- OPEN STREET VIEW moves beside the coverage hint: at the default panel
width the PROVIDERS row has no room for it, so it wrapped onto a line of
its own.
- The panel gate's corner drag names what lay under the pointer when it
fails.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The Street View step's cleanup switched the Street View chip off after the
step had left Google 3D, where the chip is hidden by design. The click then
threw "failed to find element", replacing the step's real result (here, a
Google key refusal) with a selector error. Return to Google 3D first, then
switch the chips and restore the original map.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Doctor and dev-fresh told every user without an OpenAI key that voice could
use ChatGPT OAuth, whether or not Codex was installed. Say voice is off
until a key is added or a Codex sign-in is selected, and mark that path
experimental.
The login child inherited the whole server environment, including
OPENAI_API_KEY and every other provider credential. Pass only what Codex
needs to find its home, open a browser and reach OpenAI through a proxy.
Voice-card captions from main priced transcription usage into the API-key
dollar meter. In a ChatGPT OAuth session that could trip the $2 warning or
$5 cap while the readout said COST UNKNOWN. Count them with the session's
other OAuth usage instead.
Maintainer follow-up to #966 review:
- A failed panorama-details lookup now waits DETAILS_RETRY_COOLDOWN_MS
(5 s) per panorama before a view event can retry it, so dragging the
view after a failure no longer issues a lookup on every pov change. A
success clears the pause; one panorama's failure does not delay another.
- SECURITY.md notes that allowing maps.googleapis.com in script-src also
trusts its JSONP endpoints, an accepted limitation alongside the
existing 'unsafe-eval' allowance.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
kapturner