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  follow-ups and never lets a superseded query replace follow-up memory.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

+15
View File
@@ -10,6 +10,7 @@ import {
OPENAI_REALTIME_REASONING_DEFAULT,
OPENAI_REALTIME_CONTEXT_TOKENS_DEFAULT,
OPENAI_REALTIME_CONTEXT_RETENTION_DEFAULT,
OPENAI_REALTIME_TRANSCRIBE_MODEL_DEFAULT,
} from './constants.js';
import { realtimeInstructions } from './instructions.js';
import { GEV_REALTIME_TOOLS } from './tools.js';
@@ -88,6 +89,14 @@ function createRealtimeTokenHandler({
OPENAI_REALTIME_CONTEXT_RETENTION_DEFAULT,
),
);
// User captions for the voice card. Billed separately from the session.
const transcribeModel =
process.env.OPENAI_REALTIME_TRANSCRIBE_MODEL ||
OPENAI_REALTIME_TRANSCRIBE_MODEL_DEFAULT;
const transcription =
transcribeModel.toLowerCase() === 'off'
? {}
: { transcription: { model: transcribeModel } };
const sessionConfig = {
session: {
type: 'realtime',
@@ -103,6 +112,7 @@ function createRealtimeTokenHandler({
audio: {
input: {
noise_reduction: { type: 'near_field' },
...transcription,
turn_detection: {
type: 'semantic_vad',
eagerness: 'low',
@@ -138,6 +148,11 @@ function createRealtimeTokenHandler({
// case where a bogus ?tier= was silently downgraded to standard.
res.setHeader('X-GEV-Voice-Tier', tier);
res.setHeader('X-GEV-Voice-Model', model);
// Captions are billed separately; the client meters them with this id.
res.setHeader(
'X-GEV-Voice-Transcribe-Model',
transcription.transcription ? transcribeModel : 'off',
);
if (requestedTier && !isKnownVoiceTier(requestedTier)) {
res.setHeader('X-GEV-Voice-Tier-Fallback', '1');
}