- The street-level plan step names its place the way show_street_level
reads its arguments: a pointed or numbered thing has no place yet, and
"this spot" needs both coordinates.
- The view-centre error message is removed: the Street Level layer always
gives its own translated text for an empty search, so it never showed.
"the view centre" is one exported constant for both voice modules.
- The orbital site status uses the word of the FOCUS card in es, fr, it,
pt and he, and the Wildlife chip agrees with "avistamientos" and
"avvistamenti".
- CURRENTLY OVER is the noun for an overflight in de, uk, it and pt,
shorter than before, so the place name keeps its room on a phone.
- The glossaries record the new terms; I18N.md words two rules more
exactly.
The last translation review left four kinds of problem it could only
reword around in the catalogs. They are fixed in the code now, and the
catalogs go back to natural wording.
- Counts shown with plain t() are tn() plurals: the wildlife count, the
Data Library "Showing {shown} of {total}", the News time window (which
offers 1 h) and the GhostMaps event age (which offers 1 day). The
English keys stay; the catalogs get real forms, as in "Última 1 h" and
"Últimas 24 h", or "1 Tag" and "7 Tage".
- The bridge radar range and ring spacing are tn() plurals too. The
spacing is fractional at the short ranges, and Intl.PluralRules gives
0.75 and 1.5 their own form: ru "0,75 МОР. МИЛИ", "12 МОР. МИЛЬ".
- Street level no longer passes "the view centre" or "this spot" as
{place}: each case has a message of its own, so a preposition no
longer meets a phrase in the wrong case. The progress line had the
same leak and showed the English phrase.
- Latest, FOCUS, CONFIRMED and "{label} · COURSE {heading}°" are split
with a context where one key served two places: the News heading, the
orbital site status, the Wildlife chip and a ship's course.
"CURRENTLY OVER" ended in a preposition that a country name in the
nominative cannot follow in uk, de, it and pt. It reads "under the
satellite" there now, as ru already did.
I18N.md gains the two rules behind these fixes.
Reviews the 474 strings that 27 pull requests added since the last
review, in all twelve catalogs, and fixes what the review confirmed:
wrong agreement when a placeholder takes a neutral value, plural-less
counts that read "1 sightings", terms that differ from the glossaries,
and typography the language guides set. Also applies most of the
findings earlier reviews left open.
zh-Hant is converted from zh-CN again. The new strings had been
translated separately with Taiwan vocabulary. The zh-Hant guide lists
two more characters OpenCC picks wrongly.
Three tests expected the old German and Spanish wording.
The voice card showed English for show_street_level failures the catalog
did not know: a missing key, a map or source that could not be switched on,
the layer, invalid coordinates, an imagery fallback and the filter clause.
They now translate like the other action errors (exact messages and
patterns in speech.js), in all twelve languages; the spoken line stays
English, and the filter clause names the filters as their chips do.
Review fixes for gekh's #1113 round:
- a link runs show_street_level with origin 'programmatic', so its source
switch and layer are not saved as the recipient's preferences;
- with no place named, the action looks under the screen centre
(aim 'screen-centre', armed false), keeps the sky refusal and leaves a
link's armed point alone;
- an armed OPEN aimed at the sky opens nothing instead of the far link point;
- an imagery-stage failure is spoken as an error, missing imagery as missing
(street.nearestOutcome); the card says "the view centre" in the interface
language;
- every photo the action opens reveals its panel, and revealPanel scrolls the
panel into view in the phone sheet;
- embed: EXPAND survives embed's own clean view, the lifted photo hides in
cockpit, recording and scene playback, and a failed photo step closes the
earlier photo before the view flies;
- links pasted into an open tab run their street-level step after the
restore settles; a superseded start link does not;
- Mapillary's nearest search gives up after 12 s, body included.
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.
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.
select_nearest_aircraft passes altitudeM: null for an airborne contact
without a barometric altitude (military rows from adsb.lol can lack
alt_baro). The speech builder coerced it with Number(), and Number(null)
is 0, so the reply said "..., 0 feet." and the voice card showed
"Altitude 0 feet". Read the altitude only when it is a finite number,
so an unknown one is left out as the builder already intends.
- turn named places and mapped landmarks into reusable bounded areas
- count loaded records inside resolved, drawn and annotated geometry
- find recent imagery for an area, the view or the pointer snapshot
- keep osm place search behind an operator-configured overpass endpoint
- preserve cancellation, ownership and follow-up memory across every action
- cover geometry, provider lifecycle, cache isolation and voice routing
keep contacts counts, rows, scope and provenance on one retained snapshot
report capped cohorts as lower bounds
separate sibling tool cancellation from newer user turns
align track_entity validation across runtime and benchmark paths
- Voice layer manifest (src/voice/layerManifest.js) generates the layer enums,
aliases and analyst field hints; a test fails when a catalog layer has no entry.
- analyst_query validates fields, operators, units, values and scopes, counts up
to 250,000 loaded records per layer, reports complete/partial/unanswered, tells
LAYER_OFF / NOT_READY / FEED_UNAVAILABLE apart from zero, refuses mismatched
follow-ups and never lets a superseded query replace follow-up memory. Seven
more layers expose analyst records.
- Tool results carry a code-built say line and a display card; the voice card in
the mic control shows captions, plan steps and the result
(src/voice/resultDisplay.js is the one adapter). Slow tools narrate their
current step out of band; holding Space while the assistant talks cuts it off.
Turn spans go to the debug log; caption transcription is metered; the log
omits what was said unless GEV_VOICE_LOG_CONTENT=1.
- Point and ask: the Space-down pointer snapshot resolves "this", "that one" and
"around here" through 'pointer' targets, and "the second one" through
referent:n from the last numbered list.
- Prompt states each policy once (instructions + tools 48.5k -> 46.7k chars).
- Voice benchmark harness (scripts/qa-voice-bench.mjs, scripts/voice-bench/)
shares its phrase table and grading with scripts/qa-voice-routing.mjs.