The voice answer waited 25 s for /api/osm/features, while the server let
Overpass run for 25 s and itself waited up to 40 s per endpoint and 45 s
in all. When the browser gave up first it closed the request, the route
cancelled the upstream query, and nothing was cached, so the OSM_TIMEOUT
retry started the same query from scratch and timed out again.
The server now tells Overpass 20 s, waits 22 s per endpoint and 22 s in
all, which ends inside the client's 25 s. A test pins that ordering.
A reply stopped partway through a round now keeps the actions that ran,
answering the calls it never reached as skipped, so the history stays
paired for both providers. A failed typed turn no longer resets the
pointer, referents and follow-up memory, since its conversation continues.
The server now tells a rejected key (401, or Gemini's 400 with reason
API_KEY_INVALID, matched exactly), denied access (403) and an account out
of credit (OpenAI's 429 insufficient_quota) apart; each cools down for a
minute and gets its own message in the chat.
#1308 moved /api/overpass to the keyless cross-site gate: foreign origins
and cross-site fetches are still refused, but forwarding headers are not,
so the map works behind a reverse proxy. Voice place search spends the same
operator Overpass, so it now uses the same gate; with the stricter one,
place search failed behind a proxy while the map kept working. The prompt
lint note records the merged size, 55,370 with an operator Overpass.
Picks up #1308 (Overpass admission), #114 (Sea Marks and Ski Pistes),
#1206 (radio country coverage) and #1435. Resolutions:
- SECURITY.md: main's rewritten cross-site paragraph, with voice place
search listed beside /api/overpass among the keyless map routes.
- scripts/format-scope.json: both sides. A merge on main dropped 97 entries
(being restored there); keeping them, this PR's 12 and main's 2 new ones
keeps every file in format:check. All 1,222 paths exist.
- Translation catalogs: main's two new keys and this PR's 47, no overlap.
- Layer count 36 (33 + OSM Places + Sea Marks + Ski Pistes); both schema
pins re-derived, the first-run length is exactly 30,919 + 292 + 107.
Typed chat sessions no longer end on the voice inactivity timeout, and a
failed or stopped reply keeps the conversation: Stop cancels only the
reply in progress, and map actions that already ran stay in the model's
context. Each upstream request and map action gets its own deadline, old
tool results shrink instead of the history overflowing, and step limits,
a missing or rejected key, the OAuth lock and throttling each get their
own message.
- Server: OpenAI chat is off under GEV_PREFER_CODEX_OAUTH; chat throttles
default to the provider's throttle; replies carry API_KEY_MISSING,
API_KEY_REJECTED, CODEX_OAUTH_REQUIRED and RATE_LIMITED codes; upstream
Retry-After passes through, and a rejected key (Gemini's invalid-key 400
included) cools down for a minute
- Pinokio forwards the chat throttle settings
- UI: the reply reveal publishes only the message list; the border orbit
runs only while busy; layers attach with a button or Enter; the log is
aria-busy while a reply types out; auto-scroll respects a reader who
scrolled up; Escape closes the attach menu; the chat gear says Settings;
a note says chat bills the OpenAI key when voice uses ChatGPT sign-in
- Layout does less work while panel height animates; unused blurs removed
- Gemini voice start ends chat; the voice preview clears on every new
session
- Tests for the route gates, client limits, attachments and the fixes
Review fixes for #1506.
- Ottawa: use the documented camera list and the city's Traffic Camera
Images API. The pack is skipped with no request unless
OTTAWA_CAMERA_CERTIFICATE is set; the certificate is added only at
fetch time inside the pinned /opendata/ path, and each camera is
fetched at most once a minute through the shared frame cache.
- Québec 511 video is opt-in (CCTV_QUEBEC_VIDEO_ENABLED=1) and loads
only after the viewer presses Play; selection, auto-hop, scenes and
voice no longer contact Québec 511. Credit splits locations (MTQ,
CC BY 4.0) from video (Québec 511). Interrupted play() is no longer
reported as an unavailable source; the per-frame listener runs only
while a Québec camera is selected and the canvas rect is cached.
- Catalog cap: reserved Québec and World slots come only from capacity
above the 4,000 shared slots, so a lower CCTV_MAX_SOURCES thins every
pack instead of starving the existing ones. dev-fresh and
.env.example default to 7,516.
- Throttled frames (World, Ottawa): inside the refresh floor with no
cached frame the proxy serves the placeholder, never Street View; a
failed fetch backs off for a minute; the cache is bounded by bytes.
- MJPEG first-frame extraction runs only for sources registered as
MJPEG (Taiwan Highway Bureau), caps a frame at 2 MB, reads in linear
time, accepts a CRLF preamble and a boundary that already starts with
"--", and ends a part without Content-Length at the next delimiter.
THB hosts are registered in the redirect pin list.
- Tests: World loader policy, Québec loader and opt-in, proxy cooldown
without Street View, Ottawa signing, cap reservations at 7,516, 4,000
and 48, overlay Play gate, refresh, visibility and teardown. Guard
tests check the 4,000 shared slots again.
- Remove the unused standalone Québec preview module and public page.
- Docs: DATA_SOURCES, SECURITY, CURRENT-STATE, CCTV_INTEGRATION, README,
.env.example, changelog fragment renamed to its full scope.
show_street_level (voice and typed) flies to a place and opens its nearest
street-level photo from Google Street View, Mapillary, KartaView or
Panoramax; Street View switches to Google 3D first. The view model gains a
street_level field, share links carry sl=<source>,<lat>,<lon>, and
show_in_gods_eye_view can open a photo. A plain sl=google link only arms the
point and waits for OPEN, so it never bills a key without a click.
Embed pages show an open photo in an overlay that Esc closes. On-screen
strings are translated into es, ru, zh-CN and zh-Hant. Covered by unit tests
and three new steps in the Street Level browser suite.
Tightening the AREAS rule dropped "to outline it (draw=true)", and on Gemini
"Outline the state of Texas" stopped passing draw: none of 5 runs set it, so
the model drew the state again with annotate_map. With the clause back all 4
runs called resolve_area with draw. The prompt is 55,263 characters with an
operator Overpass, ~240 under the 55,500 ceiling.
From review:
- An answer the view picked ("Springfield" over Illinois) is reused or
shared only within the same ~2° view cell, so asking again elsewhere no
longer gets the first state's area.
- Admin names compare with accents folded on Latin letters only: "Quebec"
matches "Québec" and "Cordoba" asks between Argentina and Spain, while a
Thai or Devanagari mark still tells names apart.
- A Natural Earth name used by several features is one place with every
part when they touch (the Great Rift Valley, the Lau Group across 180°)
and a question when they are far apart (Flores); natural places can now
be candidates.
- The AREAS rule and the area tool descriptions are tighter, leaving a
~270-character margin under the 55,500 prompt ceiling, and resolve_area
no longer names osm_query when that tool is not offered.
From review:
- /api/osm/features now refuses cross-site browser requests like the other
quota-bearing routes, so a page elsewhere cannot spend the operator's
Overpass. SECURITY.md lists the route, its limits and its disk cache.
- One client runs one search at a time and two run at once in all, instead
of one global slot any single client could hold.
- A search gets 45 s across every configured endpoint in turn, answered 504
when it runs out.
- A cache write no longer waits on the disk prune; prune() waits until the
cache is idle for callers that need it.
- A zoom during a pivoting tilt now scales the range the tilt holds,
as it already did for orbit, instead of being undone on the next
frame while reporting success. The zoom limits no longer move a view
the wrong way: a zoom in leaves a view closer than 80 m in place, and
a zoom out leaves an orbit beyond 5,000 km where it is.
- A grab, touch or wheel on the map while a voice fly-to is in flight
drops the motion queued behind it. Cesium lands the flight anyway,
so the orbit or tilt would otherwise take the camera back. Chained
fly-tos in one reply keep the earliest snapshot.
- A pivot tilt toward the horizon stops lowering the camera at 150 m
above the pivot and tilts the rest in place, so a close view no
longer ends inside tiles or a hillside.
- pitchDeg is an absolute angle. A direction that contradicts it is
refused, judged from the pitch the swing starts at, and the tool
description says so.
The /api/osm/features cache key was the query alone, kept for a day, so after
an operator repointed OVERPASS_UPSTREAMS the old instance's answers were still
served. The key now includes the configured-upstream tag, as main's other
Overpass caches check it, so a new instance is always asked.
Every name ending in "Zone" was kept without a level and refused as a retired
administrative identity. That is right for Nepal, which replaced its zones in
2015, but zones are current units elsewhere: Ethiopia's "Gamo Zone" was
refused. The selector now keeps the verified level unless display_name places
the zone in Nepal or names no country, and the client no longer reads "Zone"
as a retired level when the source verified one.
A voice fly-to frames a city top-down, at pitch -90. The tilt clamp
rejected any step that left the pitch below -89, so the first frame of
"tilt up" stopped the motion and "tilt down" was refused: from the view
voice lands on, no tilt ever moved the camera. A bare "tilt the camera"
also failed because tilt required a direction.
- Clamp tilt only in the direction of travel, so -90 can tilt up.
- Swing the camera about the ground under the crosshair, so the place
stays centred and the heading is kept. A tracked view still tilts in
place, and a pivot tilt yields when tracking takes the camera.
- Add an optional move_camera pitchDeg for "tilt 45 degrees"; a tilt
with neither angle nor direction settles at -45 (oblique view).
- Re-derive the voice tool schema pins for the new argument and wording,
and keep the Realtime prompt under its 52,000-character budget.
One provenanceLine() builds STATE | text | LAYERS for the HUD and
get_hud_caption. hasWord is one lookaround regex, capAtWord one grapheme
pass. Feed-state helpers are imported from layerSnapshot.js directly, and the
kept-whole sanitizer cases are one table.
Both stack parents (#837, #842) are on main, so the voice geometry branch
now builds on main directly. Conflict resolutions keep main's behaviour and
add the geometry work on top: osm_query gating on every provider, restored
admin identity for Nominatim outlines, merged analyst follow-up memory, and
the cancelled Globe View reset instruction ahead of the generic cancellation.
Resolve the CURRENT-STATE.md layer table by keeping main's realigned
table and adding this layer's row. Main added an Arabic locale, so
translate the 12 layer strings and list ar.json in the drone-airspace
package.
Starting the upstream timeout after the outbound queue let a request run
up to about 19 s, past the Directions layer's 15 s; the budget again runs
from arrival, queue wait included, and a budget spent before the request
leaves the queue answers 429 "routing busy" rather than an outage.
plan_route threw "Routing HTTP 502" for an outage, and voice passes only
the message on; it now throws "Routing service unavailable (HTTP 502)".
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A building word ending the ask ('… Chapel', '… Library', '… Hall') now
passes the last-word rule only when the result is one a building ask
would accept, so a campus, zoo or district whose name merely starts the
ask (Duke University, Old Town) no longer answers it.
- A building word ("Burj Khalifa Tower", "Sagrada Familia Church") passes
the last-word rule on a building-like result; enclosures stay out.
- The anchor chooses only among the best-named outlines, so a worse name
never wins by containment alone.
- An ask without a quadrant takes only the quadrant of the matching street
nearest the point; a leading quadrant ("SE Main St") may go unsaid too.
- On an unbuffered tile edge a fragment shows as an edge along the
boundary, so a whole building touching it at a corner is not cut.
- Drop two orphaned comments and correct the place-name key wording.
- One place-name key for server and tiles, so "St. Paul" and "Saint Paul"
agree in Nominatim name checks.
- Tile street names: a lone letter is a direction only beside another name
word ("E Street" is not "East Street"), and a trailing quadrant ("E St NW")
may go unsaid in the ask.
- A building cut at the tile edge that is nearer the point than any whole
one leaves the point instead of a farther neighbour standing in; an edge
without a tile buffer counts as a cut, so a cut area never answers grounds.
- A batch building install re-judges only results that installed; a late
timeout during caption conversion keeps the pending point drawn; a failed
plain recaption keeps the old mark addressable.
- An interrupted Nominatim dispatch is reclaimed by the next lock owner ten
minutes after the start its marker names; node:sqlite loads on first use
without its experimental warning.
- PHOTON_URL accepts only https, which the page's connect-src allows.
- Shared OpenFreeMap tiles decode each layer when a consumer first selects
it; outline selection reuses the shared ring area and simplification, and
ground probes the shared ring centroid; the grounds word list lives once.
When the OSRM routing service failed (a 5xx, a 403 block, a 404), answered
with a body that was not JSON, or did not answer at all (timeout, network
error), /api/route returned HTTP 200 with 'no route found' or 'route proxy
error'. plan_route turned that into invalid_arguments ("No walk route from
...: no route found"), blaming the caller for an outage, and the directions
layer reported "No route found between A and B".
The proxy now answers all of these with 502 {ok:false, error:'routing
service unavailable'}: a rejected upstream fetch, any failure status but
400, a non-JSON content type, and a body read that fails or does not parse.
An exception in the proxy's own code after that point is still a 200 'route
proxy error'. createRouteService throws for a 5xx before reading the body's
ok:false, so plan_route reports unavailable, and the directions source
treats any 5xx as "Routing unavailable (HTTP n)", also when the failure
page in front of the proxy is not JSON. OSRM's own refusal (a 400 such as
NoRoute, or a 200 without a route) still reads "no route found".
The 12 s upstream timeout now starts when the request leaves rather than
before it waits for an outbound slot, so a wait in the proxy's own queue
(up to about 8 s at the 1 s spacing) is not reported as an outage.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A street ribbon that holds neither its ring mean nor the anchor is centred
on the street's nearest point when the street runs within 100 m.
Nominatim outline selection ranks an exact name, then one holding every
asked word, then an overlap, and refuses an overlap whose name lacks the
ask's last word unless that word is in the result's address or type, so
"Central Park Zoo" is no longer Central Park. The server caches every
accepted outline and answers each ask with the one holding its resolved
anchor (sent as alat/alon), read across the antimeridian.
A duplicate mark that fails to clean up no longer tears down the new
outline it duplicates.
Chat sits next to the mic in the command dock. Replies are silent and never
request the microphone; the first message connects with the selected provider
and runs the same map actions as voice through each provider's text API
(OpenAI OPENAI_CHAT_MODEL, Gemini GEMINI_CHAT_MODEL via generateContent, which
never starts Live). Only one mode runs at a time.
- Chat panel rendered by the voice control presenter: conversation, recent and
quick-command suggestions, Up/Down history, attachments (map-layer
references or small UTF-8 text files, sent only with the message)
- Chat and voice panels sit above the dock or dock into the right rail's
column; minimized bars sit centred above the dock, and height animates on
swaps
- A centred panel steps down to its bar for the Location and Visual Presets
trays without moving focus, rests above an open tray, and docks right while
expanded over one; it returns to the centre once the trays close
- Minimized voice previews the last sentence from each side; the empty Voice
Preview suggests things to say
- Settings opened from chat keep chat's frame and show the shared AI provider
and quick commands; voice Settings show Back to voice
- Closing chat ends a typed session; the voice panel's close turns voice off
- The dock widens its voice control for the chat button and restyles its
wings, which show icons only at 900px and below; assistant panels share the
left panels' glass, and Cyber keeps its own surfaces
- Server text-chat handlers with body, time and per-IP limits
(GEV_RATELIMIT_*_CHAT_PER_MIN), documented in SECURITY.md; tests and all
eleven locales
Co-authored-by: Darren Dignam <[email protected]>
Han and kana have no word edges, so a CJK forbidden word now counts anywhere
in the reply. The proxy no longer counts words: the model reads Rue de Rivoli
as one, so only the 80-character word-boundary cap remains.
Also restore src/voice/gevActions.test.mjs, reformatted by mistake in 818d2c82.
Sameh Khamis