@ai-ecoverse/pi-audio 1.16.0 → 1.18.0

package/README.md

@@ -27,6 +27,8 @@ pi-audio does not apply that patch. Hosts such as slicc-agent record monkeypatch
 
 `@ai-ecoverse/pi-audio/elevenlabs` registers ElevenLabs speech adapters when `registerElevenLabs()` is called. `scribe_v2` posts multipart audio to `/speech-to-text`. `scribe_v2_realtime` streams base64 audio on `wss://…/speech-to-text/realtime` and commits turns with `turn_prediction` (`turnDetection: 'semantic'`). `eleven_flash_v2_5` and `eleven_multilingual_v2` post JSON to `/text-to-speech/{voice}/with-timestamps` for a string and stream text on `stream-input` for an async iterable, with character alignment. `eleven_v3` and `eleven_v4` stay on the HTTP path (`textStreaming: false`). Scribe prices are $0.22 per hour for batch and $0.39 per hour for realtime.
 
+`calibrateEcho(out, mic, probe, durationMs, signal)` plays a probe and aligns the mic with `AudioOut.reference`. `delayMs` is the lag. `cancellationDb` is how far the leaked copy sits below the reference, and `residualDb` is that leak. A quiet leak recommends `interrupt`, a moderate one `onWords`, and a loud one `headphones`. `chooseCapture` keeps `none` and `wasm`, and moves `browser` to `wasm` when cancellation is under 12 dB. `micOpen('browser')` asks for `echoCancellation` and `playback: 'rtc-loopback'`. `wasm` and `none` turn browser cancellation off and use `webaudio`. `assertNodeEcho` rejects `browser` on Node. `openEchoMic` leaves `browser` and `none` audio as captured. `wasm` runs an injected `EchoCanceller` against `reference`. The homescoop `@ai-ecoverse/wasm-*` echo-canceller package does not exist yet, so this package does not import one. `voiceTurns` passes `options.echo` through `openEchoMic` before `listen`.
+
 `followBed(bed, soundscape)` cues a `BedChannel` from `VoicePhase`. A loop waits `graceMs` before it starts, so a short gap stays silent. `keep` leaves the current loop running across chained tools. Speech ducks the bed to `duckUnderSpeech`; listening ducks it to `duckUnderUser`. `voiceTurns` does this when `options.soundscape` is set and `out` is an `AudioOut` (`speech`, `bed`, `reference`, `playback`). A plain `AudioSink` still receives only the spoken audio.
 
 `voiceTurns` captions include `agent_spoken`. `upTo` follows `AudioSink` playback: per character when the chunk has alignment, and evenly across the sentence when it does not. `captionsFromHistory(entries)` rebuilds user captions from entries whose `data.source` is `voice`, and agent captions from the assistant text. A `voice.truncate` entry adds `agent_spoken` and `agent_cut` at `heardUpTo`, so the rest of the original text stays available as unheard.
@@ -48,6 +50,8 @@ pi-audio does not apply that patch. Hosts such as slicc-agent record monkeypatch
 
 `mintRealtimeToken(model, constraints, options)` returns `{ value, expiresAt, model }` and calls `options.fetch` (or `globalThis.fetch`). OpenAI `POST /v1/realtime/client_secrets` sends a realtime session with instructions, voice, and tools. xAI `POST /v1/realtime/client_secrets` sends only `expires_after.seconds` of 300. Gemini `POST /v1beta/auth_tokens` locks `models/<id>`, instructions, voice, and tools in `liveConnectConstraints` and returns `name`, which the live adapter sends as `access_token`. ElevenLabs `GET /v1/convai/conversation/get-signed-url` takes `agent_id` (or `GET /conversation/token` when the only transport is WebRTC) and expires in 15 minutes. Fish `POST /v1/agent/sessions` sends `agent_id` plus `overrides.system_prompt` and `overrides.voice_id`. `value` is the LiveKit participant token. `expiresAt` is milliseconds. A numeric `expires_at` below `1e12` is unix seconds.
 
+`@ai-ecoverse/pi-audio/elevenlabs-agents` registers `elevenlabs-agent`, a hosted cascade (`provisioning: "agent"`, `truncation: "server"`, `liveToolUpdate: false`). `provisionAgent` creates or updates the agent and reuses it when the spec hash matches. pi tools become client tools. A string `llm` is a vendor model name. A `CustomLlm` sets `llm` to `custom-llm` and sends `custom_llm.url`, `model_id`, `api_type`, and `request_headers` (ElevenLabs `api_key` is a workspace secret reference, not a raw token). `connectRealtime` uses a `RealtimeToken` from `mintRealtimeToken` (`get-signed-url`) when one is passed. Otherwise an API key provisions the agent and mints the signed URL. Silent `inject` sends `contextual_update` with `context_id` `pi-audio`. `close()` prices `audio.inputSeconds` from the wall-clock session at $0.08 per minute.
+
 `talkerTools` is `ask_agent` and `read`. pi-audio defines `askAgent` (a pi `Tool` whose parameters are `request` and optional `context`) and the `Thinker` interface. The host implements `Thinker` and supplies `read`. `bindThinker` runs direct tools with `directTimeoutMs` and truncates their text to `directMaxChars`. On a blocking `ask_agent` it answers with `ack` immediately and injects the result later. On an async call it `respond`s with the result and `options.result`. Progress is injected with `options.progress`. `tool_call_cancelled` aborts the thinker. The binding also yields `thinker_progress` / `thinker_result` captions and `thinking` / `tool` / `idle` phases.
 
 ## Development

package/dist/echo.js

@@ -0,0 +1,157 @@
+import { chunkSeconds } from "./audio.js";
+import { VoiceRequestError } from "./errors.js";
+const floor = 1e-8;
+function micOpen(mode, useAll) {
+  if (mode === "browser") {
+    let echoCancellation = true;
+    if (useAll) echoCancellation = "all";
+    return { echoCancellation, playback: "rtc-loopback" };
+  }
+  return { echoCancellation: false, playback: "webaudio" };
+}
+function echoRuntime(hasMediaDevices) {
+  if (hasMediaDevices) return "browser";
+  return "node";
+}
+function assertNodeEcho(mode, runtime) {
+  if (runtime === "node" && mode === "browser") {
+    throw new VoiceRequestError("error", "Node echo control is wasm or none");
+  }
+}
+function chooseCapture(tried, report) {
+  if (tried === "none") return "none";
+  if (tried === "wasm") return "wasm";
+  if (report.cancellationDb < 12) return "wasm";
+  return "browser";
+}
+function openEchoMic(control, mic) {
+  if (control.mode === "none" || control.mode === "browser") return mic;
+  const canceller = control.canceller;
+  if (canceller === void 0) {
+    throw new VoiceRequestError("error", "wasm echo control needs a canceller");
+  }
+  return cancelStream(mic, control.reference, canceller);
+}
+function measureEcho(reference, mic, sampleRate) {
+  if (reference.length === 0) return quiet();
+  if (mic.length === 0) return quiet();
+  let rate = sampleRate;
+  if (rate <= 0) rate = 1;
+  const maxLag = Math.min(mic.length - 1, Math.max(0, Math.round(rate * 0.25)));
+  let bestLag = 0;
+  let best = -1;
+  let bestGain = 0;
+  for (let lag = 0; lag <= maxLag; lag += 1) {
+    const scored = scoreLag(reference, mic, lag);
+    if (scored.score > best) {
+      best = scored.score;
+      bestLag = lag;
+      bestGain = scored.gain;
+    }
+  }
+  const leak = Math.abs(bestGain);
+  const residualDb = amplitudeDb(leak);
+  return {
+    delayMs: bestLag / rate * 1e3,
+    cancellationDb: -residualDb,
+    residualDb,
+    recommendation: recommend(residualDb)
+  };
+}
+async function calibrateEcho(out, mic, probe, durationMs, signal) {
+  if (signal.aborted) throw new VoiceRequestError("aborted", "aborted");
+  out.bed.play(probe, 0);
+  const rate = { value: 0 };
+  try {
+    const heard = readSamples(mic, durationMs, signal, rate);
+    const reference = readSamples(out.reference, durationMs, signal, rate);
+    const pair = await Promise.all([heard, reference]);
+    let used = rate.value;
+    if (used <= 0) used = 1;
+    return measureEcho(pair[1], pair[0], used);
+  } finally {
+    out.bed.stop(0);
+  }
+}
+function quiet() {
+  return { delayMs: 0, cancellationDb: 80, residualDb: -80, recommendation: "interrupt" };
+}
+function recommend(residualDb) {
+  if (residualDb <= -30) return "interrupt";
+  if (residualDb <= -12) return "onWords";
+  return "headphones";
+}
+function scoreLag(reference, mic, lag) {
+  let dot = 0;
+  let energy = 0;
+  const length = Math.min(reference.length, mic.length - lag);
+  for (let index = 0; index < length; index += 1) {
+    const ref = reference[index];
+    const heard = mic[index + lag];
+    dot += ref * heard;
+    energy += ref * ref;
+  }
+  if (energy <= floor) return { score: -1, gain: 0 };
+  return { score: Math.abs(dot), gain: dot / energy };
+}
+function amplitudeDb(amplitude) {
+  let safe = amplitude;
+  if (safe <= floor) safe = floor;
+  return 20 * Math.log10(safe);
+}
+async function* cancelStream(mic, reference, canceller) {
+  const refs = reference[Symbol.asyncIterator]();
+  for await (const chunk of mic) {
+    const next = await refs.next();
+    if (next.done) {
+      yield chunk;
+      continue;
+    }
+    yield canceller.cancel(chunk, next.value);
+  }
+}
+async function readSamples(source, durationMs, signal, rate) {
+  const samples = [];
+  let heardMs = 0;
+  for await (const chunk of source) {
+    if (signal.aborted) throw new VoiceRequestError("aborted", "aborted");
+    if (rate.value === 0) rate.value = chunk.format.sampleRate;
+    for (const sample of samplesOf(chunk)) samples.push(sample);
+    heardMs += chunkSeconds(chunk) * 1e3;
+    if (heardMs >= durationMs) break;
+  }
+  return Float32Array.from(samples);
+}
+function samplesOf(chunk) {
+  if (chunk.format.encoding === "pcm16") return pcm16Samples(chunk.data);
+  if (chunk.format.encoding === "pcmf32") return pcmf32Samples(chunk.data);
+  return [];
+}
+function pcm16Samples(data) {
+  const view = new DataView(data.buffer, data.byteOffset, data.byteLength);
+  const samples = [];
+  const count = Math.floor(data.byteLength / 2);
+  for (let index = 0; index < count; index += 1) {
+    samples.push(view.getInt16(index * 2, true) / 32768);
+  }
+  return samples;
+}
+function pcmf32Samples(data) {
+  const view = new DataView(data.buffer, data.byteOffset, data.byteLength);
+  const samples = [];
+  const count = Math.floor(data.byteLength / 4);
+  for (let index = 0; index < count; index += 1) {
+    samples.push(view.getFloat32(index * 4, true));
+  }
+  return samples;
+}
+export {
+  assertNodeEcho,
+  calibrateEcho,
+  chooseCapture,
+  echoRuntime,
+  measureEcho,
+  micOpen,
+  openEchoMic
+};
+//# sourceMappingURL=echo.js.map

package/dist/echo.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../src/echo.ts"],
+  "sourcesContent": ["import { type AudioChunk, type AudioSource, chunkSeconds } from './audio.ts';\nimport { VoiceRequestError } from './errors.ts';\nimport type { AudioOut, Cue } from './soundscape.ts';\nimport type { EchoRecommendation } from './turns.ts';\n\nexport type EchoMode = 'browser' | 'wasm' | 'none';\n\nexport interface EchoCanceller {\n  cancel(mic: AudioChunk, reference: AudioChunk): AudioChunk;\n}\n\nexport interface EchoControl {\n  mode: EchoMode;\n  reference: AudioSource;\n  canceller?: EchoCanceller;\n}\n\nexport interface EchoReport {\n  delayMs: number;\n  cancellationDb: number;\n  residualDb: number;\n  recommendation: EchoRecommendation;\n}\n\nexport interface MicOpen {\n  echoCancellation: boolean | 'all';\n  playback: 'webaudio' | 'rtc-loopback';\n}\n\nconst floor = 1e-8;\n\nexport function micOpen(mode: EchoMode, useAll: boolean): MicOpen {\n  if (mode === 'browser') {\n    let echoCancellation: boolean | 'all' = true;\n    if (useAll) echoCancellation = 'all';\n    return { echoCancellation, playback: 'rtc-loopback' };\n  }\n  return { echoCancellation: false, playback: 'webaudio' };\n}\n\nexport function echoRuntime(hasMediaDevices: boolean): 'node' | 'browser' {\n  if (hasMediaDevices) return 'browser';\n  return 'node';\n}\n\nexport function assertNodeEcho(mode: EchoMode, runtime: 'node' | 'browser'): void {\n  if (runtime === 'node' && mode === 'browser') {\n    throw new VoiceRequestError('error', 'Node echo control is wasm or none');\n  }\n}\n\nexport function chooseCapture(tried: EchoMode, report: EchoReport): EchoMode {\n  if (tried === 'none') return 'none';\n  if (tried === 'wasm') return 'wasm';\n  if (report.cancellationDb < 12) return 'wasm';\n  return 'browser';\n}\n\nexport function openEchoMic(control: EchoControl, mic: AudioSource): AudioSource {\n  if (control.mode === 'none' || control.mode === 'browser') return mic;\n  const canceller = control.canceller;\n  if (canceller === undefined) {\n    throw new VoiceRequestError('error', 'wasm echo control needs a canceller');\n  }\n  return cancelStream(mic, control.reference, canceller);\n}\n\nexport function measureEcho(\n  reference: Float32Array,\n  mic: Float32Array,\n  sampleRate: number\n): EchoReport {\n  if (reference.length === 0) return quiet();\n  if (mic.length === 0) return quiet();\n  let rate = sampleRate;\n  if (rate <= 0) rate = 1;\n  const maxLag = Math.min(mic.length - 1, Math.max(0, Math.round(rate * 0.25)));\n  let bestLag = 0;\n  let best = -1;\n  let bestGain = 0;\n  for (let lag = 0; lag <= maxLag; lag += 1) {\n    const scored = scoreLag(reference, mic, lag);\n    if (scored.score > best) {\n      best = scored.score;\n      bestLag = lag;\n      bestGain = scored.gain;\n    }\n  }\n  const leak = Math.abs(bestGain);\n  const residualDb = amplitudeDb(leak);\n  return {\n    delayMs: (bestLag / rate) * 1000,\n    cancellationDb: -residualDb,\n    residualDb,\n    recommendation: recommend(residualDb),\n  };\n}\n\nexport async function calibrateEcho(\n  out: AudioOut,\n  mic: AudioSource,\n  probe: Cue,\n  durationMs: number,\n  signal: AbortSignal\n): Promise<EchoReport> {\n  if (signal.aborted) throw new VoiceRequestError('aborted', 'aborted');\n  out.bed.play(probe, 0);\n  const rate = { value: 0 };\n  try {\n    const heard = readSamples(mic, durationMs, signal, rate);\n    const reference = readSamples(out.reference, durationMs, signal, rate);\n    const pair = await Promise.all([heard, reference]);\n    let used = rate.value;\n    if (used <= 0) used = 1;\n    return measureEcho(pair[1], pair[0], used);\n  } finally {\n    out.bed.stop(0);\n  }\n}\n\nfunction quiet(): EchoReport {\n  return { delayMs: 0, cancellationDb: 80, residualDb: -80, recommendation: 'interrupt' };\n}\n\nfunction recommend(residualDb: number): EchoRecommendation {\n  if (residualDb <= -30) return 'interrupt';\n  if (residualDb <= -12) return 'onWords';\n  return 'headphones';\n}\n\nfunction scoreLag(\n  reference: Float32Array,\n  mic: Float32Array,\n  lag: number\n): { score: number; gain: number } {\n  let dot = 0;\n  let energy = 0;\n  const length = Math.min(reference.length, mic.length - lag);\n  for (let index = 0; index < length; index += 1) {\n    const ref = reference[index] as number;\n    const heard = mic[index + lag] as number;\n    dot += ref * heard;\n    energy += ref * ref;\n  }\n  if (energy <= floor) return { score: -1, gain: 0 };\n  return { score: Math.abs(dot), gain: dot / energy };\n}\n\nfunction amplitudeDb(amplitude: number): number {\n  let safe = amplitude;\n  if (safe <= floor) safe = floor;\n  return 20 * Math.log10(safe);\n}\n\nasync function* cancelStream(\n  mic: AudioSource,\n  reference: AudioSource,\n  canceller: EchoCanceller\n): AudioSource {\n  const refs = reference[Symbol.asyncIterator]();\n  for await (const chunk of mic) {\n    const next = await refs.next();\n    if (next.done) {\n      yield chunk;\n      continue;\n    }\n    yield canceller.cancel(chunk, next.value);\n  }\n}\n\nasync function readSamples(\n  source: AudioSource,\n  durationMs: number,\n  signal: AbortSignal,\n  rate: { value: number }\n): Promise<Float32Array> {\n  const samples: number[] = [];\n  let heardMs = 0;\n  for await (const chunk of source) {\n    if (signal.aborted) throw new VoiceRequestError('aborted', 'aborted');\n    if (rate.value === 0) rate.value = chunk.format.sampleRate;\n    for (const sample of samplesOf(chunk)) samples.push(sample);\n    heardMs += chunkSeconds(chunk) * 1000;\n    if (heardMs >= durationMs) break;\n  }\n  return Float32Array.from(samples);\n}\n\nfunction samplesOf(chunk: AudioChunk): number[] {\n  if (chunk.format.encoding === 'pcm16') return pcm16Samples(chunk.data);\n  if (chunk.format.encoding === 'pcmf32') return pcmf32Samples(chunk.data);\n  return [];\n}\n\nfunction pcm16Samples(data: Uint8Array): number[] {\n  const view = new DataView(data.buffer, data.byteOffset, data.byteLength);\n  const samples: number[] = [];\n  const count = Math.floor(data.byteLength / 2);\n  for (let index = 0; index < count; index += 1) {\n    samples.push(view.getInt16(index * 2, true) / 32768);\n  }\n  return samples;\n}\n\nfunction pcmf32Samples(data: Uint8Array): number[] {\n  const view = new DataView(data.buffer, data.byteOffset, data.byteLength);\n  const samples: number[] = [];\n  const count = Math.floor(data.byteLength / 4);\n  for (let index = 0; index < count; index += 1) {\n    samples.push(view.getFloat32(index * 4, true));\n  }\n  return samples;\n}\n"],
+  "mappings": "AAAA,SAA4C,oBAAoB;AAChE,SAAS,yBAAyB;AA4BlC,MAAM,QAAQ;AAEP,SAAS,QAAQ,MAAgB,QAA0B;AAChE,MAAI,SAAS,WAAW;AACtB,QAAI,mBAAoC;AACxC,QAAI,OAAQ,oBAAmB;AAC/B,WAAO,EAAE,kBAAkB,UAAU,eAAe;AAAA,EACtD;AACA,SAAO,EAAE,kBAAkB,OAAO,UAAU,WAAW;AACzD;AAEO,SAAS,YAAY,iBAA8C;AACxE,MAAI,gBAAiB,QAAO;AAC5B,SAAO;AACT;AAEO,SAAS,eAAe,MAAgB,SAAmC;AAChF,MAAI,YAAY,UAAU,SAAS,WAAW;AAC5C,UAAM,IAAI,kBAAkB,SAAS,mCAAmC;AAAA,EAC1E;AACF;AAEO,SAAS,cAAc,OAAiB,QAA8B;AAC3E,MAAI,UAAU,OAAQ,QAAO;AAC7B,MAAI,UAAU,OAAQ,QAAO;AAC7B,MAAI,OAAO,iBAAiB,GAAI,QAAO;AACvC,SAAO;AACT;AAEO,SAAS,YAAY,SAAsB,KAA+B;AAC/E,MAAI,QAAQ,SAAS,UAAU,QAAQ,SAAS,UAAW,QAAO;AAClE,QAAM,YAAY,QAAQ;AAC1B,MAAI,cAAc,QAAW;AAC3B,UAAM,IAAI,kBAAkB,SAAS,qCAAqC;AAAA,EAC5E;AACA,SAAO,aAAa,KAAK,QAAQ,WAAW,SAAS;AACvD;AAEO,SAAS,YACd,WACA,KACA,YACY;AACZ,MAAI,UAAU,WAAW,EAAG,QAAO,MAAM;AACzC,MAAI,IAAI,WAAW,EAAG,QAAO,MAAM;AACnC,MAAI,OAAO;AACX,MAAI,QAAQ,EAAG,QAAO;AACtB,QAAM,SAAS,KAAK,IAAI,IAAI,SAAS,GAAG,KAAK,IAAI,GAAG,KAAK,MAAM,OAAO,IAAI,CAAC,CAAC;AAC5E,MAAI,UAAU;AACd,MAAI,OAAO;AACX,MAAI,WAAW;AACf,WAAS,MAAM,GAAG,OAAO,QAAQ,OAAO,GAAG;AACzC,UAAM,SAAS,SAAS,WAAW,KAAK,GAAG;AAC3C,QAAI,OAAO,QAAQ,MAAM;AACvB,aAAO,OAAO;AACd,gBAAU;AACV,iBAAW,OAAO;AAAA,IACpB;AAAA,EACF;AACA,QAAM,OAAO,KAAK,IAAI,QAAQ;AAC9B,QAAM,aAAa,YAAY,IAAI;AACnC,SAAO;AAAA,IACL,SAAU,UAAU,OAAQ;AAAA,IAC5B,gBAAgB,CAAC;AAAA,IACjB;AAAA,IACA,gBAAgB,UAAU,UAAU;AAAA,EACtC;AACF;AAEA,eAAsB,cACpB,KACA,KACA,OACA,YACA,QACqB;AACrB,MAAI,OAAO,QAAS,OAAM,IAAI,kBAAkB,WAAW,SAAS;AACpE,MAAI,IAAI,KAAK,OAAO,CAAC;AACrB,QAAM,OAAO,EAAE,OAAO,EAAE;AACxB,MAAI;AACF,UAAM,QAAQ,YAAY,KAAK,YAAY,QAAQ,IAAI;AACvD,UAAM,YAAY,YAAY,IAAI,WAAW,YAAY,QAAQ,IAAI;AACrE,UAAM,OAAO,MAAM,QAAQ,IAAI,CAAC,OAAO,SAAS,CAAC;AACjD,QAAI,OAAO,KAAK;AAChB,QAAI,QAAQ,EAAG,QAAO;AACtB,WAAO,YAAY,KAAK,CAAC,GAAG,KAAK,CAAC,GAAG,IAAI;AAAA,EAC3C,UAAE;AACA,QAAI,IAAI,KAAK,CAAC;AAAA,EAChB;AACF;AAEA,SAAS,QAAoB;AAC3B,SAAO,EAAE,SAAS,GAAG,gBAAgB,IAAI,YAAY,KAAK,gBAAgB,YAAY;AACxF;AAEA,SAAS,UAAU,YAAwC;AACzD,MAAI,cAAc,IAAK,QAAO;AAC9B,MAAI,cAAc,IAAK,QAAO;AAC9B,SAAO;AACT;AAEA,SAAS,SACP,WACA,KACA,KACiC;AACjC,MAAI,MAAM;AACV,MAAI,SAAS;AACb,QAAM,SAAS,KAAK,IAAI,UAAU,QAAQ,IAAI,SAAS,GAAG;AAC1D,WAAS,QAAQ,GAAG,QAAQ,QAAQ,SAAS,GAAG;AAC9C,UAAM,MAAM,UAAU,KAAK;AAC3B,UAAM,QAAQ,IAAI,QAAQ,GAAG;AAC7B,WAAO,MAAM;AACb,cAAU,MAAM;AAAA,EAClB;AACA,MAAI,UAAU,MAAO,QAAO,EAAE,OAAO,IAAI,MAAM,EAAE;AACjD,SAAO,EAAE,OAAO,KAAK,IAAI,GAAG,GAAG,MAAM,MAAM,OAAO;AACpD;AAEA,SAAS,YAAY,WAA2B;AAC9C,MAAI,OAAO;AACX,MAAI,QAAQ,MAAO,QAAO;AAC1B,SAAO,KAAK,KAAK,MAAM,IAAI;AAC7B;AAEA,gBAAgB,aACd,KACA,WACA,WACa;AACb,QAAM,OAAO,UAAU,OAAO,aAAa,EAAE;AAC7C,mBAAiB,SAAS,KAAK;AAC7B,UAAM,OAAO,MAAM,KAAK,KAAK;AAC7B,QAAI,KAAK,MAAM;AACb,YAAM;AACN;AAAA,IACF;AACA,UAAM,UAAU,OAAO,OAAO,KAAK,KAAK;AAAA,EAC1C;AACF;AAEA,eAAe,YACb,QACA,YACA,QACA,MACuB;AACvB,QAAM,UAAoB,CAAC;AAC3B,MAAI,UAAU;AACd,mBAAiB,SAAS,QAAQ;AAChC,QAAI,OAAO,QAAS,OAAM,IAAI,kBAAkB,WAAW,SAAS;AACpE,QAAI,KAAK,UAAU,EAAG,MAAK,QAAQ,MAAM,OAAO;AAChD,eAAW,UAAU,UAAU,KAAK,EAAG,SAAQ,KAAK,MAAM;AAC1D,eAAW,aAAa,KAAK,IAAI;AACjC,QAAI,WAAW,WAAY;AAAA,EAC7B;AACA,SAAO,aAAa,KAAK,OAAO;AAClC;AAEA,SAAS,UAAU,OAA6B;AAC9C,MAAI,MAAM,OAAO,aAAa,QAAS,QAAO,aAAa,MAAM,IAAI;AACrE,MAAI,MAAM,OAAO,aAAa,SAAU,QAAO,cAAc,MAAM,IAAI;AACvE,SAAO,CAAC;AACV;AAEA,SAAS,aAAa,MAA4B;AAChD,QAAM,OAAO,IAAI,SAAS,KAAK,QAAQ,KAAK,YAAY,KAAK,UAAU;AACvE,QAAM,UAAoB,CAAC;AAC3B,QAAM,QAAQ,KAAK,MAAM,KAAK,aAAa,CAAC;AAC5C,WAAS,QAAQ,GAAG,QAAQ,OAAO,SAAS,GAAG;AAC7C,YAAQ,KAAK,KAAK,SAAS,QAAQ,GAAG,IAAI,IAAI,KAAK;AAAA,EACrD;AACA,SAAO;AACT;AAEA,SAAS,cAAc,MAA4B;AACjD,QAAM,OAAO,IAAI,SAAS,KAAK,QAAQ,KAAK,YAAY,KAAK,UAAU;AACvE,QAAM,UAAoB,CAAC;AAC3B,QAAM,QAAQ,KAAK,MAAM,KAAK,aAAa,CAAC;AAC5C,WAAS,QAAQ,GAAG,QAAQ,OAAO,SAAS,GAAG;AAC7C,YAAQ,KAAK,KAAK,WAAW,QAAQ,GAAG,IAAI,CAAC;AAAA,EAC/C;AACA,SAAO;AACT;",
+  "names": []
+}

package/dist/elevenlabs-agents/events.js

@@ -0,0 +1,172 @@
+import { blankUsage } from "../blank-usage.js";
+import { decodeAudioBase64 } from "../convert.js";
+function none() {
+  return { events: [], replies: [] };
+}
+function isRecord(value) {
+  return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+function text(value) {
+  return typeof value === "string" ? value : "";
+}
+function transcript(itemId, value, final) {
+  return { type: "output_transcript", itemId, text: value, final };
+}
+function done(itemId, interrupted) {
+  return { type: "response_done", itemId, interrupted, usage: blankUsage() };
+}
+function numbers(value) {
+  if (!Array.isArray(value) || value.some((item) => typeof item !== "number")) return void 0;
+  return value;
+}
+function alignmentOf(value) {
+  if (!isRecord(value)) return void 0;
+  const startMs = numbers(value.char_start_times_ms);
+  const durationMs = numbers(value.char_durations_ms);
+  if (startMs === void 0 || durationMs === void 0) return void 0;
+  return { startMs, durationMs };
+}
+function pong(_turn, message) {
+  const nested = isRecord(message.ping_event) ? message.ping_event : void 0;
+  const raw = nested === void 0 ? message.event_id : nested.event_id;
+  const eventId = typeof raw === "number" ? raw : 0;
+  return { events: [], replies: [{ type: "pong", event_id: eventId }] };
+}
+function metadata(_turn, message) {
+  const event = isRecord(message.conversation_initiation_metadata_event) ? message.conversation_initiation_metadata_event : void 0;
+  const sessionId = text(event?.conversation_id);
+  return {
+    events: [
+      { type: "session_start", sessionId: sessionId.length > 0 ? sessionId : "conversation" }
+    ],
+    replies: []
+  };
+}
+function userTranscript(_turn, message) {
+  const event = isRecord(message.user_transcription_event) ? message.user_transcription_event : void 0;
+  return {
+    events: [
+      {
+        type: "input_transcript",
+        itemId: "user",
+        text: text(event?.user_transcript),
+        final: true
+      }
+    ],
+    replies: []
+  };
+}
+function chatPart(_turn, message) {
+  const part = isRecord(message.text_response_part) ? message.text_response_part : void 0;
+  if (part === void 0) return none();
+  const kind = text(part.type);
+  if (kind === "start" || kind.length === 0) return none();
+  const itemId = text(part.response_id);
+  return {
+    events: [transcript(itemId.length > 0 ? itemId : "agent", text(part.text), kind === "stop")],
+    replies: []
+  };
+}
+function agentResponse(_turn, message) {
+  const event = isRecord(message.agent_response_event) ? message.agent_response_event : void 0;
+  return { events: [transcript("agent", text(event?.agent_response), true)], replies: [] };
+}
+function correction(_turn, message) {
+  const event = isRecord(message.agent_response_correction_event) ? message.agent_response_correction_event : void 0;
+  return {
+    events: [transcript("agent", text(event?.corrected_agent_response), true)],
+    replies: []
+  };
+}
+function audio(turn, message, format) {
+  const event = isRecord(message.audio_event) ? message.audio_event : void 0;
+  const eventId = typeof event?.event_id === "number" ? event.event_id : 0;
+  if (turn.dropAudio || eventId <= turn.lastInterruptId) return none();
+  const alignment = alignmentOf(event?.alignment);
+  const chars = isRecord(event?.alignment) ? event.alignment.chars : void 0;
+  const end = Array.isArray(chars) ? chars.length : 0;
+  const itemId = String(eventId);
+  const first = alignment?.startMs[0];
+  const spoken = {
+    type: "output_audio",
+    itemId,
+    chunk: {
+      data: decodeAudioBase64(text(event?.audio_base_64)),
+      format,
+      timestamp: typeof first === "number" ? first : 0
+    }
+  };
+  if (alignment !== void 0) spoken.alignment = alignment;
+  if (end > 0) spoken.span = { itemId, start: 0, end };
+  return { events: [spoken], replies: [] };
+}
+function interruption(turn, message) {
+  const event = isRecord(message.interruption_event) ? message.interruption_event : void 0;
+  const eventId = typeof event?.event_id === "number" ? event.event_id : turn.lastInterruptId;
+  turn.lastInterruptId = eventId;
+  turn.vad = true;
+  return { events: [{ type: "input_speech_start" }, done(String(eventId), true)], replies: [] };
+}
+function vad(turn, message) {
+  const event = isRecord(message.vad_score_event) ? message.vad_score_event : void 0;
+  const score = typeof event?.vad_score === "number" ? event.vad_score : 0;
+  const speaking = score > 0.5;
+  if (!speaking) {
+    turn.vad = false;
+    return none();
+  }
+  if (turn.vad) return none();
+  turn.vad = true;
+  return { events: [{ type: "input_speech_start" }], replies: [] };
+}
+function toolCall(_turn, message) {
+  const event = isRecord(message.client_tool_call) ? message.client_tool_call : void 0;
+  const name = text(event?.tool_name);
+  const id = text(event?.tool_call_id);
+  if (name.length === 0 || id.length === 0) return none();
+  const parameters = isRecord(event?.parameters) ? event.parameters : {};
+  const call = {
+    type: "toolCall",
+    id,
+    name,
+    arguments: parameters
+  };
+  return { events: [{ type: "tool_call", call, blocking: true }], replies: [] };
+}
+function complete(turn, _message) {
+  turn.dropAudio = false;
+  return { events: [done("agent", false)], replies: [] };
+}
+function clientError(_turn, message) {
+  const event = isRecord(message.client_error_event) ? message.client_error_event : void 0;
+  const reason = text(event?.message);
+  return {
+    events: [{ type: "error", error: reason.length > 0 ? reason : "client error", fatal: false }],
+    replies: []
+  };
+}
+const handlers = {
+  ping: pong,
+  conversation_initiation_metadata: metadata,
+  user_transcript: userTranscript,
+  agent_chat_response_part: chatPart,
+  agent_response: agentResponse,
+  agent_response_correction: correction,
+  audio,
+  interruption,
+  vad_score: vad,
+  client_tool_call: toolCall,
+  agent_response_complete: complete,
+  client_error: clientError
+};
+function ingestElevenLabs(turn, message, format) {
+  const type = message.type;
+  if (typeof type !== "string") return none();
+  const handler = handlers[type];
+  if (handler === void 0) return none();
+  return handler(turn, message, format);
+}
+export {
+  ingestElevenLabs
+};
+//# sourceMappingURL=events.js.map

package/dist/elevenlabs-agents/events.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/elevenlabs-agents/events.ts"],
+  "sourcesContent": ["import type { JsonObject, ToolCall } from '@earendil-works/pi-ai';\nimport type { AudioFormat, CharAlignment } from '../audio.ts';\nimport { blankUsage } from '../blank-usage.ts';\nimport { decodeAudioBase64 } from '../convert.ts';\nimport type { RealtimeEvent } from '../realtime.ts';\n\nexport interface TalkTurn {\n  vad: boolean;\n  dropAudio: boolean;\n  lastInterruptId: number;\n}\n\nexport interface Ingested {\n  events: RealtimeEvent[];\n  replies: JsonObject[];\n}\n\nfunction none(): Ingested {\n  return { events: [], replies: [] };\n}\n\nfunction isRecord(value: unknown): value is Record<string, unknown> {\n  return typeof value === 'object' && value !== null && !Array.isArray(value);\n}\n\nfunction text(value: unknown): string {\n  return typeof value === 'string' ? value : '';\n}\n\nfunction transcript(itemId: string, value: string, final: boolean): RealtimeEvent {\n  return { type: 'output_transcript', itemId, text: value, final };\n}\n\nfunction done(itemId: string, interrupted: boolean): RealtimeEvent {\n  return { type: 'response_done', itemId, interrupted, usage: blankUsage() };\n}\n\nfunction numbers(value: unknown): number[] | undefined {\n  if (!Array.isArray(value) || value.some((item) => typeof item !== 'number')) return undefined;\n  return value as number[];\n}\n\nfunction alignmentOf(value: unknown): CharAlignment | undefined {\n  if (!isRecord(value)) return undefined;\n  const startMs = numbers(value.char_start_times_ms);\n  const durationMs = numbers(value.char_durations_ms);\n  if (startMs === undefined || durationMs === undefined) return undefined;\n  return { startMs, durationMs };\n}\n\nfunction pong(_turn: TalkTurn, message: Record<string, unknown>): Ingested {\n  const nested = isRecord(message.ping_event) ? message.ping_event : undefined;\n  const raw = nested === undefined ? message.event_id : nested.event_id;\n  const eventId = typeof raw === 'number' ? raw : 0;\n  return { events: [], replies: [{ type: 'pong', event_id: eventId }] };\n}\n\nfunction metadata(_turn: TalkTurn, message: Record<string, unknown>): Ingested {\n  const event = isRecord(message.conversation_initiation_metadata_event)\n    ? message.conversation_initiation_metadata_event\n    : undefined;\n  const sessionId = text(event?.conversation_id);\n  return {\n    events: [\n      { type: 'session_start', sessionId: sessionId.length > 0 ? sessionId : 'conversation' },\n    ],\n    replies: [],\n  };\n}\n\nfunction userTranscript(_turn: TalkTurn, message: Record<string, unknown>): Ingested {\n  const event = isRecord(message.user_transcription_event)\n    ? message.user_transcription_event\n    : undefined;\n  return {\n    events: [\n      {\n        type: 'input_transcript',\n        itemId: 'user',\n        text: text(event?.user_transcript),\n        final: true,\n      },\n    ],\n    replies: [],\n  };\n}\n\nfunction chatPart(_turn: TalkTurn, message: Record<string, unknown>): Ingested {\n  const part = isRecord(message.text_response_part) ? message.text_response_part : undefined;\n  if (part === undefined) return none();\n  const kind = text(part.type);\n  if (kind === 'start' || kind.length === 0) return none();\n  const itemId = text(part.response_id);\n  return {\n    events: [transcript(itemId.length > 0 ? itemId : 'agent', text(part.text), kind === 'stop')],\n    replies: [],\n  };\n}\n\nfunction agentResponse(_turn: TalkTurn, message: Record<string, unknown>): Ingested {\n  const event = isRecord(message.agent_response_event) ? message.agent_response_event : undefined;\n  return { events: [transcript('agent', text(event?.agent_response), true)], replies: [] };\n}\n\nfunction correction(_turn: TalkTurn, message: Record<string, unknown>): Ingested {\n  const event = isRecord(message.agent_response_correction_event)\n    ? message.agent_response_correction_event\n    : undefined;\n  return {\n    events: [transcript('agent', text(event?.corrected_agent_response), true)],\n    replies: [],\n  };\n}\n\nfunction audio(turn: TalkTurn, message: Record<string, unknown>, format: AudioFormat): Ingested {\n  const event = isRecord(message.audio_event) ? message.audio_event : undefined;\n  const eventId = typeof event?.event_id === 'number' ? event.event_id : 0;\n  if (turn.dropAudio || eventId <= turn.lastInterruptId) return none();\n  const alignment = alignmentOf(event?.alignment);\n  const chars = isRecord(event?.alignment) ? event.alignment.chars : undefined;\n  const end = Array.isArray(chars) ? chars.length : 0;\n  const itemId = String(eventId);\n  const first = alignment?.startMs[0];\n  const spoken: RealtimeEvent = {\n    type: 'output_audio',\n    itemId,\n    chunk: {\n      data: decodeAudioBase64(text(event?.audio_base_64)),\n      format,\n      timestamp: typeof first === 'number' ? first : 0,\n    },\n  };\n  if (alignment !== undefined) spoken.alignment = alignment;\n  if (end > 0) spoken.span = { itemId, start: 0, end };\n  return { events: [spoken], replies: [] };\n}\n\nfunction interruption(turn: TalkTurn, message: Record<string, unknown>): Ingested {\n  const event = isRecord(message.interruption_event) ? message.interruption_event : undefined;\n  const eventId = typeof event?.event_id === 'number' ? event.event_id : turn.lastInterruptId;\n  turn.lastInterruptId = eventId;\n  turn.vad = true;\n  return { events: [{ type: 'input_speech_start' }, done(String(eventId), true)], replies: [] };\n}\n\nfunction vad(turn: TalkTurn, message: Record<string, unknown>): Ingested {\n  const event = isRecord(message.vad_score_event) ? message.vad_score_event : undefined;\n  const score = typeof event?.vad_score === 'number' ? event.vad_score : 0;\n  const speaking = score > 0.5;\n  if (!speaking) {\n    turn.vad = false;\n    return none();\n  }\n  if (turn.vad) return none();\n  turn.vad = true;\n  return { events: [{ type: 'input_speech_start' }], replies: [] };\n}\n\nfunction toolCall(_turn: TalkTurn, message: Record<string, unknown>): Ingested {\n  const event = isRecord(message.client_tool_call) ? message.client_tool_call : undefined;\n  const name = text(event?.tool_name);\n  const id = text(event?.tool_call_id);\n  if (name.length === 0 || id.length === 0) return none();\n  const parameters = isRecord(event?.parameters) ? event.parameters : {};\n  const call: ToolCall = {\n    type: 'toolCall',\n    id,\n    name,\n    arguments: parameters as JsonObject,\n  };\n  return { events: [{ type: 'tool_call', call, blocking: true }], replies: [] };\n}\n\nfunction complete(turn: TalkTurn, _message: Record<string, unknown>): Ingested {\n  turn.dropAudio = false;\n  return { events: [done('agent', false)], replies: [] };\n}\n\nfunction clientError(_turn: TalkTurn, message: Record<string, unknown>): Ingested {\n  const event = isRecord(message.client_error_event) ? message.client_error_event : undefined;\n  const reason = text(event?.message);\n  return {\n    events: [{ type: 'error', error: reason.length > 0 ? reason : 'client error', fatal: false }],\n    replies: [],\n  };\n}\n\nconst handlers: Record<\n  string,\n  (turn: TalkTurn, message: Record<string, unknown>, format: AudioFormat) => Ingested\n> = {\n  ping: pong,\n  conversation_initiation_metadata: metadata,\n  user_transcript: userTranscript,\n  agent_chat_response_part: chatPart,\n  agent_response: agentResponse,\n  agent_response_correction: correction,\n  audio,\n  interruption,\n  vad_score: vad,\n  client_tool_call: toolCall,\n  agent_response_complete: complete,\n  client_error: clientError,\n};\n\nexport function ingestElevenLabs(\n  turn: TalkTurn,\n  message: Record<string, unknown>,\n  format: AudioFormat\n): Ingested {\n  const type = message.type;\n  if (typeof type !== 'string') return none();\n  const handler = handlers[type];\n  if (handler === undefined) return none();\n  return handler(turn, message, format);\n}\n"],
+  "mappings": "AAEA,SAAS,kBAAkB;AAC3B,SAAS,yBAAyB;AAclC,SAAS,OAAiB;AACxB,SAAO,EAAE,QAAQ,CAAC,GAAG,SAAS,CAAC,EAAE;AACnC;AAEA,SAAS,SAAS,OAAkD;AAClE,SAAO,OAAO,UAAU,YAAY,UAAU,QAAQ,CAAC,MAAM,QAAQ,KAAK;AAC5E;AAEA,SAAS,KAAK,OAAwB;AACpC,SAAO,OAAO,UAAU,WAAW,QAAQ;AAC7C;AAEA,SAAS,WAAW,QAAgB,OAAe,OAA+B;AAChF,SAAO,EAAE,MAAM,qBAAqB,QAAQ,MAAM,OAAO,MAAM;AACjE;AAEA,SAAS,KAAK,QAAgB,aAAqC;AACjE,SAAO,EAAE,MAAM,iBAAiB,QAAQ,aAAa,OAAO,WAAW,EAAE;AAC3E;AAEA,SAAS,QAAQ,OAAsC;AACrD,MAAI,CAAC,MAAM,QAAQ,KAAK,KAAK,MAAM,KAAK,CAAC,SAAS,OAAO,SAAS,QAAQ,EAAG,QAAO;AACpF,SAAO;AACT;AAEA,SAAS,YAAY,OAA2C;AAC9D,MAAI,CAAC,SAAS,KAAK,EAAG,QAAO;AAC7B,QAAM,UAAU,QAAQ,MAAM,mBAAmB;AACjD,QAAM,aAAa,QAAQ,MAAM,iBAAiB;AAClD,MAAI,YAAY,UAAa,eAAe,OAAW,QAAO;AAC9D,SAAO,EAAE,SAAS,WAAW;AAC/B;AAEA,SAAS,KAAK,OAAiB,SAA4C;AACzE,QAAM,SAAS,SAAS,QAAQ,UAAU,IAAI,QAAQ,aAAa;AACnE,QAAM,MAAM,WAAW,SAAY,QAAQ,WAAW,OAAO;AAC7D,QAAM,UAAU,OAAO,QAAQ,WAAW,MAAM;AAChD,SAAO,EAAE,QAAQ,CAAC,GAAG,SAAS,CAAC,EAAE,MAAM,QAAQ,UAAU,QAAQ,CAAC,EAAE;AACtE;AAEA,SAAS,SAAS,OAAiB,SAA4C;AAC7E,QAAM,QAAQ,SAAS,QAAQ,sCAAsC,IACjE,QAAQ,yCACR;AACJ,QAAM,YAAY,KAAK,OAAO,eAAe;AAC7C,SAAO;AAAA,IACL,QAAQ;AAAA,MACN,EAAE,MAAM,iBAAiB,WAAW,UAAU,SAAS,IAAI,YAAY,eAAe;AAAA,IACxF;AAAA,IACA,SAAS,CAAC;AAAA,EACZ;AACF;AAEA,SAAS,eAAe,OAAiB,SAA4C;AACnF,QAAM,QAAQ,SAAS,QAAQ,wBAAwB,IACnD,QAAQ,2BACR;AACJ,SAAO;AAAA,IACL,QAAQ;AAAA,MACN;AAAA,QACE,MAAM;AAAA,QACN,QAAQ;AAAA,QACR,MAAM,KAAK,OAAO,eAAe;AAAA,QACjC,OAAO;AAAA,MACT;AAAA,IACF;AAAA,IACA,SAAS,CAAC;AAAA,EACZ;AACF;AAEA,SAAS,SAAS,OAAiB,SAA4C;AAC7E,QAAM,OAAO,SAAS,QAAQ,kBAAkB,IAAI,QAAQ,qBAAqB;AACjF,MAAI,SAAS,OAAW,QAAO,KAAK;AACpC,QAAM,OAAO,KAAK,KAAK,IAAI;AAC3B,MAAI,SAAS,WAAW,KAAK,WAAW,EAAG,QAAO,KAAK;AACvD,QAAM,SAAS,KAAK,KAAK,WAAW;AACpC,SAAO;AAAA,IACL,QAAQ,CAAC,WAAW,OAAO,SAAS,IAAI,SAAS,SAAS,KAAK,KAAK,IAAI,GAAG,SAAS,MAAM,CAAC;AAAA,IAC3F,SAAS,CAAC;AAAA,EACZ;AACF;AAEA,SAAS,cAAc,OAAiB,SAA4C;AAClF,QAAM,QAAQ,SAAS,QAAQ,oBAAoB,IAAI,QAAQ,uBAAuB;AACtF,SAAO,EAAE,QAAQ,CAAC,WAAW,SAAS,KAAK,OAAO,cAAc,GAAG,IAAI,CAAC,GAAG,SAAS,CAAC,EAAE;AACzF;AAEA,SAAS,WAAW,OAAiB,SAA4C;AAC/E,QAAM,QAAQ,SAAS,QAAQ,+BAA+B,IAC1D,QAAQ,kCACR;AACJ,SAAO;AAAA,IACL,QAAQ,CAAC,WAAW,SAAS,KAAK,OAAO,wBAAwB,GAAG,IAAI,CAAC;AAAA,IACzE,SAAS,CAAC;AAAA,EACZ;AACF;AAEA,SAAS,MAAM,MAAgB,SAAkC,QAA+B;AAC9F,QAAM,QAAQ,SAAS,QAAQ,WAAW,IAAI,QAAQ,cAAc;AACpE,QAAM,UAAU,OAAO,OAAO,aAAa,WAAW,MAAM,WAAW;AACvE,MAAI,KAAK,aAAa,WAAW,KAAK,gBAAiB,QAAO,KAAK;AACnE,QAAM,YAAY,YAAY,OAAO,SAAS;AAC9C,QAAM,QAAQ,SAAS,OAAO,SAAS,IAAI,MAAM,UAAU,QAAQ;AACnE,QAAM,MAAM,MAAM,QAAQ,KAAK,IAAI,MAAM,SAAS;AAClD,QAAM,SAAS,OAAO,OAAO;AAC7B,QAAM,QAAQ,WAAW,QAAQ,CAAC;AAClC,QAAM,SAAwB;AAAA,IAC5B,MAAM;AAAA,IACN;AAAA,IACA,OAAO;AAAA,MACL,MAAM,kBAAkB,KAAK,OAAO,aAAa,CAAC;AAAA,MAClD;AAAA,MACA,WAAW,OAAO,UAAU,WAAW,QAAQ;AAAA,IACjD;AAAA,EACF;AACA,MAAI,cAAc,OAAW,QAAO,YAAY;AAChD,MAAI,MAAM,EAAG,QAAO,OAAO,EAAE,QAAQ,OAAO,GAAG,IAAI;AACnD,SAAO,EAAE,QAAQ,CAAC,MAAM,GAAG,SAAS,CAAC,EAAE;AACzC;AAEA,SAAS,aAAa,MAAgB,SAA4C;AAChF,QAAM,QAAQ,SAAS,QAAQ,kBAAkB,IAAI,QAAQ,qBAAqB;AAClF,QAAM,UAAU,OAAO,OAAO,aAAa,WAAW,MAAM,WAAW,KAAK;AAC5E,OAAK,kBAAkB;AACvB,OAAK,MAAM;AACX,SAAO,EAAE,QAAQ,CAAC,EAAE,MAAM,qBAAqB,GAAG,KAAK,OAAO,OAAO,GAAG,IAAI,CAAC,GAAG,SAAS,CAAC,EAAE;AAC9F;AAEA,SAAS,IAAI,MAAgB,SAA4C;AACvE,QAAM,QAAQ,SAAS,QAAQ,eAAe,IAAI,QAAQ,kBAAkB;AAC5E,QAAM,QAAQ,OAAO,OAAO,cAAc,WAAW,MAAM,YAAY;AACvE,QAAM,WAAW,QAAQ;AACzB,MAAI,CAAC,UAAU;AACb,SAAK,MAAM;AACX,WAAO,KAAK;AAAA,EACd;AACA,MAAI,KAAK,IAAK,QAAO,KAAK;AAC1B,OAAK,MAAM;AACX,SAAO,EAAE,QAAQ,CAAC,EAAE,MAAM,qBAAqB,CAAC,GAAG,SAAS,CAAC,EAAE;AACjE;AAEA,SAAS,SAAS,OAAiB,SAA4C;AAC7E,QAAM,QAAQ,SAAS,QAAQ,gBAAgB,IAAI,QAAQ,mBAAmB;AAC9E,QAAM,OAAO,KAAK,OAAO,SAAS;AAClC,QAAM,KAAK,KAAK,OAAO,YAAY;AACnC,MAAI,KAAK,WAAW,KAAK,GAAG,WAAW,EAAG,QAAO,KAAK;AACtD,QAAM,aAAa,SAAS,OAAO,UAAU,IAAI,MAAM,aAAa,CAAC;AACrE,QAAM,OAAiB;AAAA,IACrB,MAAM;AAAA,IACN;AAAA,IACA;AAAA,IACA,WAAW;AAAA,EACb;AACA,SAAO,EAAE,QAAQ,CAAC,EAAE,MAAM,aAAa,MAAM,UAAU,KAAK,CAAC,GAAG,SAAS,CAAC,EAAE;AAC9E;AAEA,SAAS,SAAS,MAAgB,UAA6C;AAC7E,OAAK,YAAY;AACjB,SAAO,EAAE,QAAQ,CAAC,KAAK,SAAS,KAAK,CAAC,GAAG,SAAS,CAAC,EAAE;AACvD;AAEA,SAAS,YAAY,OAAiB,SAA4C;AAChF,QAAM,QAAQ,SAAS,QAAQ,kBAAkB,IAAI,QAAQ,qBAAqB;AAClF,QAAM,SAAS,KAAK,OAAO,OAAO;AAClC,SAAO;AAAA,IACL,QAAQ,CAAC,EAAE,MAAM,SAAS,OAAO,OAAO,SAAS,IAAI,SAAS,gBAAgB,OAAO,MAAM,CAAC;AAAA,IAC5F,SAAS,CAAC;AAAA,EACZ;AACF;AAEA,MAAM,WAGF;AAAA,EACF,MAAM;AAAA,EACN,kCAAkC;AAAA,EAClC,iBAAiB;AAAA,EACjB,0BAA0B;AAAA,EAC1B,gBAAgB;AAAA,EAChB,2BAA2B;AAAA,EAC3B;AAAA,EACA;AAAA,EACA,WAAW;AAAA,EACX,kBAAkB;AAAA,EAClB,yBAAyB;AAAA,EACzB,cAAc;AAChB;AAEO,SAAS,iBACd,MACA,SACA,QACU;AACV,QAAM,OAAO,QAAQ;AACrB,MAAI,OAAO,SAAS,SAAU,QAAO,KAAK;AAC1C,QAAM,UAAU,SAAS,IAAI;AAC7B,MAAI,YAAY,OAAW,QAAO,KAAK;AACvC,SAAO,QAAQ,MAAM,SAAS,MAAM;AACtC;",
+  "names": []
+}

package/dist/elevenlabs-agents/index.js

@@ -0,0 +1,58 @@
+import { registerVoiceModel } from "../models.js";
+import { registerRealtimeApiProvider } from "../registry.js";
+import { connectElevenLabs } from "./session.js";
+const voices = [
+  { id: "21m00Tcm4TlvDq8ikWAM", name: "Rachel" },
+  { id: "EXAVITQu4vr4xnSDxMaL", name: "Sarah" },
+  { id: "pNInz6obpgDQGcFmaJgB", name: "Adam" }
+];
+function pcm16(sampleRate) {
+  return { encoding: "pcm16", sampleRate, channels: 1 };
+}
+const elevenLabsAgent = {
+  id: "elevenlabs-agent",
+  name: "ElevenLabs Agent",
+  api: "elevenlabs-agents",
+  provider: "elevenlabs",
+  baseUrl: "wss://api.elevenlabs.io/v1/convai/conversation",
+  type: "realtime",
+  transports: ["websocket"],
+  toolCalls: "blocking",
+  inputFormat: pcm16(16e3),
+  outputFormat: pcm16(16e3),
+  maxSessionMs: 6e5,
+  resumable: false,
+  ephemeralAuth: true,
+  voices,
+  cascade: { stt: "scribe_realtime", llm: "gemini-2.5-flash", tts: "eleven_flash_v2_5" },
+  provisioning: "agent",
+  delivery: { interrupt: "approximate", when_idle: "approximate", silent: "native" },
+  truncation: "server",
+  liveToolUpdate: false,
+  toolTimeoutMs: { min: 1e3, max: 12e4 },
+  cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+  audioCost: { perInputMinute: 0.08 }
+};
+function registerElevenLabsAgents(options = {}) {
+  const unregisterModel = registerVoiceModel(elevenLabsAgent);
+  const unregisterApi = registerRealtimeApiProvider({
+    api: "elevenlabs-agents",
+    connect: (model, context, connectOptions) => connectElevenLabs(model, context, connectOptions, options)
+  });
+  return () => {
+    unregisterModel();
+    unregisterApi();
+  };
+}
+import { hashHostedSpec, memoryHostedCache } from "../hosted-agent.js";
+import { agentBody, httpsOrigin, provisionAgent } from "./provision.js";
+export {
+  agentBody,
+  elevenLabsAgent,
+  hashHostedSpec,
+  httpsOrigin,
+  memoryHostedCache,
+  provisionAgent,
+  registerElevenLabsAgents
+};
+//# sourceMappingURL=index.js.map

package/dist/elevenlabs-agents/index.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../src/elevenlabs-agents/index.ts"],
+  "sourcesContent": ["import type { RealtimeModel } from '../models.ts';\nimport { registerVoiceModel } from '../models.ts';\nimport { registerRealtimeApiProvider } from '../registry.ts';\nimport type { ElevenLabsOptions } from './session.ts';\nimport { connectElevenLabs } from './session.ts';\n\nconst voices = [\n  { id: '21m00Tcm4TlvDq8ikWAM', name: 'Rachel' },\n  { id: 'EXAVITQu4vr4xnSDxMaL', name: 'Sarah' },\n  { id: 'pNInz6obpgDQGcFmaJgB', name: 'Adam' },\n];\n\nfunction pcm16(sampleRate: number) {\n  return { encoding: 'pcm16' as const, sampleRate, channels: 1 as const };\n}\n\nexport const elevenLabsAgent: RealtimeModel<'elevenlabs-agents'> = {\n  id: 'elevenlabs-agent',\n  name: 'ElevenLabs Agent',\n  api: 'elevenlabs-agents',\n  provider: 'elevenlabs',\n  baseUrl: 'wss://api.elevenlabs.io/v1/convai/conversation',\n  type: 'realtime',\n  transports: ['websocket'],\n  toolCalls: 'blocking',\n  inputFormat: pcm16(16000),\n  outputFormat: pcm16(16000),\n  maxSessionMs: 600_000,\n  resumable: false,\n  ephemeralAuth: true,\n  voices,\n  cascade: { stt: 'scribe_realtime', llm: 'gemini-2.5-flash', tts: 'eleven_flash_v2_5' },\n  provisioning: 'agent',\n  delivery: { interrupt: 'approximate', when_idle: 'approximate', silent: 'native' },\n  truncation: 'server',\n  liveToolUpdate: false,\n  toolTimeoutMs: { min: 1000, max: 120_000 },\n  cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },\n  audioCost: { perInputMinute: 0.08 },\n};\n\nexport function registerElevenLabsAgents(options: ElevenLabsOptions = {}): () => void {\n  const unregisterModel = registerVoiceModel(elevenLabsAgent);\n  const unregisterApi = registerRealtimeApiProvider({\n    api: 'elevenlabs-agents',\n    connect: (model, context, connectOptions) =>\n      connectElevenLabs(model, context, connectOptions, options),\n  });\n  return () => {\n    unregisterModel();\n    unregisterApi();\n  };\n}\n\nexport type { CustomLlm, HostedAgent, HostedAgentSpec, HostedCache } from '../hosted-agent.ts';\nexport { hashHostedSpec, memoryHostedCache } from '../hosted-agent.ts';\nexport type { ProvisionOptions } from './provision.ts';\nexport { agentBody, httpsOrigin, provisionAgent } from './provision.ts';\nexport type { ElevenLabsOptions, VoiceWireFactory } from './session.ts';\n"],
+  "mappings": "AACA,SAAS,0BAA0B;AACnC,SAAS,mCAAmC;AAE5C,SAAS,yBAAyB;AAElC,MAAM,SAAS;AAAA,EACb,EAAE,IAAI,wBAAwB,MAAM,SAAS;AAAA,EAC7C,EAAE,IAAI,wBAAwB,MAAM,QAAQ;AAAA,EAC5C,EAAE,IAAI,wBAAwB,MAAM,OAAO;AAC7C;AAEA,SAAS,MAAM,YAAoB;AACjC,SAAO,EAAE,UAAU,SAAkB,YAAY,UAAU,EAAW;AACxE;AAEO,MAAM,kBAAsD;AAAA,EACjE,IAAI;AAAA,EACJ,MAAM;AAAA,EACN,KAAK;AAAA,EACL,UAAU;AAAA,EACV,SAAS;AAAA,EACT,MAAM;AAAA,EACN,YAAY,CAAC,WAAW;AAAA,EACxB,WAAW;AAAA,EACX,aAAa,MAAM,IAAK;AAAA,EACxB,cAAc,MAAM,IAAK;AAAA,EACzB,cAAc;AAAA,EACd,WAAW;AAAA,EACX,eAAe;AAAA,EACf;AAAA,EACA,SAAS,EAAE,KAAK,mBAAmB,KAAK,oBAAoB,KAAK,oBAAoB;AAAA,EACrF,cAAc;AAAA,EACd,UAAU,EAAE,WAAW,eAAe,WAAW,eAAe,QAAQ,SAAS;AAAA,EACjF,YAAY;AAAA,EACZ,gBAAgB;AAAA,EAChB,eAAe,EAAE,KAAK,KAAM,KAAK,KAAQ;AAAA,EACzC,MAAM,EAAE,OAAO,GAAG,QAAQ,GAAG,WAAW,GAAG,YAAY,EAAE;AAAA,EACzD,WAAW,EAAE,gBAAgB,KAAK;AACpC;AAEO,SAAS,yBAAyB,UAA6B,CAAC,GAAe;AACpF,QAAM,kBAAkB,mBAAmB,eAAe;AAC1D,QAAM,gBAAgB,4BAA4B;AAAA,IAChD,KAAK;AAAA,IACL,SAAS,CAAC,OAAO,SAAS,mBACxB,kBAAkB,OAAO,SAAS,gBAAgB,OAAO;AAAA,EAC7D,CAAC;AACD,SAAO,MAAM;AACX,oBAAgB;AAChB,kBAAc;AAAA,EAChB;AACF;AAGA,SAAS,gBAAgB,yBAAyB;AAElD,SAAS,WAAW,aAAa,sBAAsB;",
+  "names": []
+}

package/dist/elevenlabs-agents/provision.js

@@ -0,0 +1,167 @@
+import { VoiceRequestError } from "../errors.js";
+import {
+  hashHostedSpec,
+  hostedCacheKey,
+  memoryHostedCache
+} from "../hosted-agent.js";
+const defaultCache = memoryHostedCache();
+const clientEvents = [
+  "conversation_initiation_metadata",
+  "ping",
+  "audio",
+  "interruption",
+  "user_transcript",
+  "agent_response",
+  "agent_response_correction",
+  "client_tool_call",
+  "vad_score",
+  "agent_chat_response_part",
+  "agent_response_complete"
+];
+function isRecord(value) {
+  return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+function asVoice(error) {
+  if (error instanceof VoiceRequestError) return error;
+  if (error instanceof Error && error.name === "AbortError") {
+    return new VoiceRequestError("aborted", "session aborted");
+  }
+  if (error instanceof Error) return new VoiceRequestError("error", error.message);
+  return new VoiceRequestError("error", String(error));
+}
+function httpsOrigin(baseUrl) {
+  const url = new URL(baseUrl);
+  let protocol = url.protocol;
+  if (protocol === "wss:") protocol = "https:";
+  if (protocol === "ws:") protocol = "http:";
+  return `${protocol}//${url.host}`;
+}
+function mergedHeaders(options, defaults) {
+  const headers = { ...defaults };
+  const extra = options.headers;
+  if (extra === void 0) return headers;
+  for (const [key, value] of Object.entries(extra)) {
+    if (value === null) delete headers[key];
+    else headers[key] = value;
+  }
+  return headers;
+}
+function apiKey(options) {
+  const key = options.apiKey;
+  if (typeof key === "string" && key.length > 0) return key;
+  throw new VoiceRequestError("error", "missing realtime credential");
+}
+function timeoutSeconds(timeoutMs) {
+  if (timeoutMs === void 0) return 20;
+  return Math.min(120, Math.max(1, Math.round(timeoutMs / 1e3)));
+}
+function clientTool(tool, timeoutMs) {
+  return {
+    type: "client",
+    name: tool.name,
+    description: tool.description,
+    expects_response: true,
+    response_timeout_secs: timeoutSeconds(timeoutMs),
+    parameters: tool.parameters
+  };
+}
+function customLlmFields(llm) {
+  const custom = {
+    url: llm.baseUrl,
+    model_id: llm.id,
+    api_type: llm.api === "openai-responses" ? "responses" : "chat_completions"
+  };
+  const headers = llm.headers;
+  if (headers !== void 0 && Object.keys(headers).length > 0) custom.request_headers = headers;
+  return { llm: "custom-llm", custom_llm: custom };
+}
+function promptFields(spec) {
+  const llm = typeof spec.llm === "string" ? { llm: spec.llm } : customLlmFields(spec.llm);
+  return {
+    prompt: spec.instructions,
+    tools: spec.tools.map((tool) => clientTool(tool, spec.clientToolTimeoutMs)),
+    ...llm
+  };
+}
+function agentBody(spec) {
+  const agent = { prompt: promptFields(spec) };
+  if (spec.firstMessage !== void 0) agent.first_message = spec.firstMessage;
+  return {
+    name: spec.name,
+    conversation_config: {
+      agent,
+      tts: { voice_id: spec.voice },
+      conversation: { client_events: clientEvents }
+    },
+    platform_settings: {
+      overrides: {
+        conversation_config_override: {
+          agent: { prompt: { prompt: true }, first_message: true },
+          tts: { voice_id: true }
+        },
+        custom_llm_extra_body: true
+      }
+    }
+  };
+}
+async function requestJson(fetchImpl, url, init) {
+  let response;
+  try {
+    response = await fetchImpl(url, init);
+  } catch (error) {
+    throw asVoice(error);
+  }
+  if (!response.ok) {
+    throw new VoiceRequestError("error", `agent request failed (${response.status})`);
+  }
+  let json;
+  try {
+    json = await response.json();
+  } catch (error) {
+    throw asVoice(error);
+  }
+  if (!isRecord(json)) throw new VoiceRequestError("error", "agent response was not an object");
+  return json;
+}
+function agentIdFrom(json, fallback) {
+  const id = json.agent_id;
+  if (typeof id === "string" && id.length > 0) return id;
+  if (fallback !== void 0) return fallback;
+  throw new VoiceRequestError("error", "agent response had no agent_id");
+}
+async function provisionAgent(model, spec, options = {}) {
+  if (options.signal?.aborted === true) throw new VoiceRequestError("aborted", "session aborted");
+  const cache = options.cache ?? defaultCache;
+  const now = options.now ?? Date.now;
+  const key = hostedCacheKey(model.provider, spec.name);
+  const hash = hashHostedSpec(spec);
+  const cached = cache.get(key);
+  if (cached !== void 0 && cached.specHash === hash) return cached;
+  const existing = cached?.agentId ?? options.agentId;
+  const origin = httpsOrigin(model.baseUrl);
+  const url = existing === void 0 ? `${origin}/v1/convai/agents/create` : `${origin}/v1/convai/agents/${encodeURIComponent(existing)}`;
+  const headers = mergedHeaders(options, {
+    "Content-Type": "application/json",
+    "xi-api-key": apiKey(options)
+  });
+  const fetchImpl = options.fetch ?? globalThis.fetch;
+  const json = await requestJson(fetchImpl, url, {
+    method: existing === void 0 ? "POST" : "PATCH",
+    headers,
+    body: JSON.stringify(agentBody(spec)),
+    signal: options.signal
+  });
+  const agent = {
+    agentId: agentIdFrom(json, existing),
+    specHash: hash,
+    publishedAt: now()
+  };
+  cache.set(key, agent);
+  return agent;
+}
+export {
+  agentBody,
+  httpsOrigin,
+  provisionAgent
+};
+//# sourceMappingURL=provision.js.map