A review of the Russian UI screenshots found ~90 places where longer words
broke the layout or text was still English. Fixed so they hold for any
language, then shortened where a control is short by design.
- Layout: label columns that size to their text (weather and wind settings,
Recent Imagery slots and opacity, style sliders, display panel), layer
list names and icons, buttons that wrap instead of overflowing (CCTV,
scenes, Bhote Koshi), voice header that always shows the cost, styled
EN | ES | RU chips, voice error tray above the dock.
- t() context: t('CLEAR', { context: 'action' }) is «СБРОСИТЬ» while the sky
stays «ЯСНО»; also used for radio categories, voice states and directions.
- Units and dates follow the locale through src/ui/unitLabels.js
(formatQuantity: localized number, translated unit, non-breaking space):
wind, pressure, cyclones, directions, CCTV, SDR gain, scene timing, flight
tooltip, Recent Imagery dates and resolution.
- Newly translated: directions turn-by-turn steps (built from the maneuver;
English matches the server word for word), HUD mode and summary, voice
card states, radio categories on the globe, CCTV status, layer source
lines, scene dialogs (Cancel, file pickers), Bhote Koshi statuses, titles,
comparison dates and caveat, Recent Imagery notices. Stored data and text
for AI tools stay English.
- es and ru: 1,885 messages each.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Mark every UI string the unmarked-text check reported (534) plus text it
cannot see (helpers, joined labels, runtime-built panels), and translate
the result to Spanish and Russian: 1,679 messages each.
- Kept on purpose, with the reason where the text lives: provider and
product names, place and camera names, codes and units, dB values, the HUD
terminal readouts, demo shot titles that saved projects look up by title,
and text for AI tools and the voice model.
- Modules shared with the server or AI tools stay free of the i18n import
(local msg); the wind reading and voice session stay English and are
translated where the page shows them.
- npm test now runs the unmarked-text check.
- Fixes on the way: a duplicated radio privacy paragraph, the contact panel
toggle saying Expand while open, the CSS filter mistaking & for CSS.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
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.