feat(voice): voice core — honest layer reach, spoken summary with on-screen card, narration, point-and-ask

- 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  follow-ups and never lets a superseded query replace follow-up memory.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.
This commit is contained in:
Bilawal Sidhu committed 2026-09-28 21:19:23 -05:00
1 parent e7707d9a0f
commit a19897ad15
97 files changed
+14343 -755

No files matched your search

+6 -4
View File
@@ -667,12 +667,14 @@ test('the voice TOOL SCHEMA matches the pinned release — the mission mapping i
// Cyber deliberately adds one layout; first-run missions still change no tools.
hudLayout.enum = hudLayout.enum.filter((layout) => layout !== 'cyber');
const block = JSON.stringify(legacyTools);
// Re-derived for the additive `local-adsb` set_layer_visibility value and
// its common-name mapping; the missions still ride existing tools.
assert.equal(block.length, 27432, 'serialized tool schema length drifted');
// Re-derived for the voice layer manifest (generated layer enums, alias and
// field hints), point-and-ask's pointer/referent arguments and the prompt
// consolidation (shorter analyst, annotate_map and ISS wording); the
// missions still ride existing tools.
assert.equal(block.length, 28800, 'serialized tool schema length drifted');
assert.equal(
crypto.createHash('sha256').update(block).digest('hex'),
'a2a4a787f4528f75b01f3f42caec636f29c37452c0d45b11f4f986171d6be57d',
'7e346dadcc6e6b520c77a5c8cae501c9ee9a2b6e6c61babbe1b0f564cd396db3',
'the first-run missions must ride EXISTING tools: no schema edit, no cache bust',
);
const instructions = fs.readFileSync(new URL('../server/providers/openai/instructions.js', import.meta.url), 'utf8');