{
  "product": "Lugemi Live",
  "model_id": "lugemi-live",
  "model_version": "local-demo-1",
  "family": "Echo + Translate + Voice",
  "note": "Incremental interpretation with learned wait/commit policy and transparent repair. States: receiving → provisional → committed → spoken; repair_required after commitment; cancelled before playback. Spoken audio is immutable. Source/target pickers cover all 344 country-pack corridors (164/206 countries). Evaluation depth varies.",
  "corridor_count": 344,
  "countries_covered": 164,
  "country_pack_total": 206,
  "apis": {
    "engine": "GET /v1/live/engine",
    "sessions": "POST /v1/live/sessions",
    "audio": "POST /v1/live/sessions/:id/audio",
    "events": "GET /v1/live/sessions/:id/events",
    "ack": "POST /v1/live/sessions/:id/ack",
    "repair": "POST /v1/live/sessions/:id/repair",
    "corridors": "GET /v1/portfolio/corridors"
  },
  "transport": "Session negotiation over HTTP; versioned events delivered via Server-Sent Events at GET /v1/live/sessions/:id/events (same event schema as the portfolio brief). A transport=websocket request is accepted for compatibility but events still stream over SSE — there is no native WebSocket upgrade in this local/demo adapter.",
  "console": "/live",
  "docs": "/docs/next-model-portfolio/03_LIVE.md"
}