@ai-ecoverse/pi-audio 1.6.0 → 1.7.0

package/README.md

@@ -31,6 +31,8 @@ pi-audio does not apply that patch. Hosts such as slicc-agent record monkeypatch
 
 `RealtimeLedger` meters one session. Input seconds come from `appendAudio`. Output seconds come from each `output_audio` event: PCM, mulaw, and alaw from the byte length, and a compressed encoding from character alignment. `sendText` and each delivered `inject` add one text message. Tool results do not. `close()` prices those meters with `priceAudio`. Sum that usage once. `response_done.usage` is the adapter's per-response figure, not a second copy of the session total.
 
+`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
 
 ```sh

package/dist/index.js

@@ -23,6 +23,7 @@ import {
   transcribe
 } from "./registry.js";
 import { SpeechEventStream } from "./speech.js";
+import { askAgent, bindThinker, talkerTools } from "./talker.js";
 import { TranscriptionEventStream } from "./transcription.js";
 import { addAudioCounters, priceAudio } from "./usage.js";
 import { energyVad } from "./vad.js";
@@ -32,6 +33,8 @@ export {
   TranscriptionEventStream,
   VoiceRequestError,
   addAudioCounters,
+  askAgent,
+  bindThinker,
   blankUsage,
   connectRealtime,
   convertAudio,
@@ -52,6 +55,7 @@ export {
   registerVoiceModel,
   resolveVoiceAuth,
   speak,
+  talkerTools,
   transcribe
 };
 //# sourceMappingURL=index.js.map

package/dist/index.js.map

@@ -1,7 +1,7 @@
 {
   "version": 3,
   "sources": ["../src/index.ts"],
-  "sourcesContent": ["export type {\n  AudioChunk,\n  AudioEncoding,\n  AudioFormat,\n  AudioSink,\n  AudioSource,\n  AudioTag,\n  CharAlignment,\n  TextSpan,\n} from './audio.ts';\nexport { resolveVoiceAuth } from './auth.ts';\nexport { blankUsage } from './blank-usage.ts';\nexport {\n  convertAudio,\n  convertAudioSource,\n  decodeAudioBase64,\n  encodeAudioBase64,\n} from './convert.ts';\nexport { VoiceRequestError } from './errors.ts';\nexport type {\n  Delivery,\n  DeliverySupport,\n  KnownRealtimeApi,\n  KnownSpeechApi,\n  KnownTranscriptionApi,\n  RealtimeApi,\n  RealtimeModel,\n  SpeechApi,\n  SpeechModel,\n  TranscriptionApi,\n  TranscriptionModel,\n  VoiceInfo,\n  VoiceModel,\n  VoiceModelType,\n} from './models.ts';\nexport {\n  getVoiceModel,\n  listVoiceModels,\n  registerVoiceModel,\n} from './models.ts';\nexport type {\n  DeliveryPlan,\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeEvent,\n  RealtimeOptions,\n  RealtimeSession,\n  RealtimeSocket,\n  RealtimeToken,\n  TruncationPlan,\n} from './realtime.ts';\nexport { planDelivery, planTruncation, RealtimeLedger, realtimeSocket } from './realtime.ts';\nexport {\n  connectRealtime,\n  listen,\n  registerRealtimeApiProvider,\n  registerSpeechApiProvider,\n  registerTranscriptionApiProvider,\n  speak,\n  transcribe,\n} from './registry.ts';\nexport type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';\nexport { SpeechEventStream } from './speech.ts';\nexport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEvent,\n  TranscriptSegment,\n} from './transcription.ts';\nexport { TranscriptionEventStream } from './transcription.ts';\nexport type {\n  AudioCost,\n  AudioCounters,\n  AudioUsage,\n  RealtimeAudioCost,\n  SpeechAudioCost,\n  TranscriptionAudioCost,\n} from './usage.ts';\nexport { addAudioCounters, priceAudio } from './usage.ts';\nexport type { VadHook, VadUpdate } from './vad.ts';\nexport { energyVad } from './vad.ts';\n"],
-  "mappings": "AAUA,SAAS,wBAAwB;AACjC,SAAS,kBAAkB;AAC3B;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAAyB;AAiBlC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAYP,SAAS,cAAc,gBAAgB,gBAAgB,sBAAsB;AAC7E;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,yBAAyB;AASlC,SAAS,gCAAgC;AASzC,SAAS,kBAAkB,kBAAkB;AAE7C,SAAS,iBAAiB;",
+  "sourcesContent": ["export type {\n  AudioChunk,\n  AudioEncoding,\n  AudioFormat,\n  AudioSink,\n  AudioSource,\n  AudioTag,\n  CharAlignment,\n  TextSpan,\n} from './audio.ts';\nexport { resolveVoiceAuth } from './auth.ts';\nexport { blankUsage } from './blank-usage.ts';\nexport {\n  convertAudio,\n  convertAudioSource,\n  decodeAudioBase64,\n  encodeAudioBase64,\n} from './convert.ts';\nexport { VoiceRequestError } from './errors.ts';\nexport type {\n  Delivery,\n  DeliverySupport,\n  KnownRealtimeApi,\n  KnownSpeechApi,\n  KnownTranscriptionApi,\n  RealtimeApi,\n  RealtimeModel,\n  SpeechApi,\n  SpeechModel,\n  TranscriptionApi,\n  TranscriptionModel,\n  VoiceInfo,\n  VoiceModel,\n  VoiceModelType,\n} from './models.ts';\nexport {\n  getVoiceModel,\n  listVoiceModels,\n  registerVoiceModel,\n} from './models.ts';\nexport type {\n  DeliveryPlan,\n  RealtimeApiProvider,\n  RealtimeContext,\n  RealtimeEvent,\n  RealtimeOptions,\n  RealtimeSession,\n  RealtimeSocket,\n  RealtimeToken,\n  TruncationPlan,\n} from './realtime.ts';\nexport { planDelivery, planTruncation, RealtimeLedger, realtimeSocket } from './realtime.ts';\nexport {\n  connectRealtime,\n  listen,\n  registerRealtimeApiProvider,\n  registerSpeechApiProvider,\n  registerTranscriptionApiProvider,\n  speak,\n  transcribe,\n} from './registry.ts';\nexport type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';\nexport { SpeechEventStream } from './speech.ts';\nexport type {\n  BindThinkerOptions,\n  BoundThinker,\n  Thinker,\n  ThinkerCaption,\n  ThinkerPhase,\n  ThinkerUpdate,\n} from './talker.ts';\nexport { askAgent, bindThinker, talkerTools } from './talker.ts';\nexport type {\n  AudioInput,\n  ListenOptions,\n  Transcript,\n  TranscriptionApiProvider,\n  TranscriptionEvent,\n  TranscriptSegment,\n} from './transcription.ts';\nexport { TranscriptionEventStream } from './transcription.ts';\nexport type {\n  AudioCost,\n  AudioCounters,\n  AudioUsage,\n  RealtimeAudioCost,\n  SpeechAudioCost,\n  TranscriptionAudioCost,\n} from './usage.ts';\nexport { addAudioCounters, priceAudio } from './usage.ts';\nexport type { VadHook, VadUpdate } from './vad.ts';\nexport { energyVad } from './vad.ts';\n"],
+  "mappings": "AAUA,SAAS,wBAAwB;AACjC,SAAS,kBAAkB;AAC3B;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAAyB;AAiBlC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAYP,SAAS,cAAc,gBAAgB,gBAAgB,sBAAsB;AAC7E;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,yBAAyB;AASlC,SAAS,UAAU,aAAa,mBAAmB;AASnD,SAAS,gCAAgC;AASzC,SAAS,kBAAkB,kBAAkB;AAE7C,SAAS,iBAAiB;",
   "names": []
 }

package/dist/talker.js

@@ -0,0 +1,192 @@
+import {
+  Type
+} from "@earendil-works/pi-ai";
+import { VoiceRequestError } from "./errors.js";
+const heardLimit = 20;
+const talkerTools = ["ask_agent", "read"];
+const askAgent = {
+  name: "ask_agent",
+  description: "Hand a request to the coding agent. Use for anything that needs files, the browser, tools or more than a few seconds of thought.",
+  parameters: Type.Object({
+    request: Type.String(),
+    context: Type.Optional(Type.String())
+  })
+};
+class PushStream {
+  #queue = [];
+  #waiters = [];
+  #closed = false;
+  push(value) {
+    if (this.#closed) return;
+    const waiter = this.#waiters.shift();
+    if (waiter) waiter({ value, done: false });
+    else this.#queue.push(value);
+  }
+  close() {
+    if (this.#closed) return;
+    this.#closed = true;
+    while (this.#waiters.length > 0) this.#waiters.shift()?.({ done: true });
+  }
+  async *[Symbol.asyncIterator]() {
+    while (true) {
+      const queued = this.#queue.shift();
+      if (queued !== void 0) {
+        yield queued;
+        continue;
+      }
+      if (this.#closed) return;
+      const next = await new Promise(
+        (resolve) => this.#waiters.push(resolve)
+      );
+      if (next.done || next.value === void 0) return;
+      yield next.value;
+    }
+  }
+}
+function send(action) {
+  try {
+    action();
+  } catch (error) {
+    if (!(error instanceof VoiceRequestError)) throw error;
+  }
+}
+function textOf(message) {
+  return message.content.filter((part) => part.type === "text").map((part) => part.text).join("\n");
+}
+function acknowledged(call, ack) {
+  return {
+    role: "toolResult",
+    toolCallId: call.id,
+    toolName: call.name,
+    content: [{ type: "text", text: ack }],
+    isError: false,
+    timestamp: Date.now()
+  };
+}
+function errorResult(call, error) {
+  return {
+    role: "toolResult",
+    toolCallId: call.id,
+    toolName: call.name,
+    content: [{ type: "text", text: error instanceof Error ? error.message : String(error) }],
+    isError: true,
+    timestamp: Date.now()
+  };
+}
+function capResult(result, maxChars) {
+  let left = maxChars;
+  const content = result.content.map((part) => {
+    if (part.type !== "text") return part;
+    if (left <= 0) return { ...part, text: "" };
+    if (part.text.length <= left) {
+      left -= part.text.length;
+      return part;
+    }
+    const text = part.text.slice(0, left);
+    left = 0;
+    return { ...part, text };
+  });
+  return { ...result, content };
+}
+function bindThinker(session, thinker, direct, options) {
+  const captions = new PushStream();
+  const phases = new PushStream();
+  const done = run(session, thinker, direct, options, captions, phases).finally(() => {
+    captions.close();
+    phases.close();
+    captions.close();
+    phases.close();
+  });
+  return { captions, phases, done };
+}
+async function run(session, thinker, direct, options, captions, phases) {
+  const heard = [];
+  const running = /* @__PURE__ */ new Map();
+  for await (const event of session) {
+    if (event.type === "input_transcript" && event.final) heard.push(`user: ${event.text}`);
+    if (event.type === "output_transcript" && event.final) heard.push(`talker: ${event.text}`);
+    if (event.type === "tool_call_cancelled") {
+      const job2 = running.get(event.callId);
+      if (!job2) continue;
+      job2.abort.abort();
+      if (!job2.answered) {
+        job2.answered = true;
+        send(() => session.respond(event.callId, errorResult(job2.call, new Error("cancelled"))));
+      }
+      continue;
+    }
+    if (event.type !== "tool_call") continue;
+    const call = event.call;
+    if (call.name !== askAgent.name) {
+      phases.push({
+        phase: "tool",
+        toolName: call.name,
+        toolCallId: call.id,
+        args: call.arguments
+      });
+      void direct(call, AbortSignal.timeout(options.directTimeoutMs)).then((result) => {
+        send(() => session.respond(call.id, capResult(result, options.directMaxChars)));
+        phases.push({ phase: "idle" });
+      }).catch((error) => {
+        send(() => session.respond(call.id, errorResult(call, error)));
+        phases.push({ phase: "idle" });
+      });
+      continue;
+    }
+    const job = { abort: new AbortController(), answered: event.blocking, call };
+    running.set(call.id, job);
+    phases.push({ phase: "thinking" });
+    if (event.blocking) send(() => session.respond(call.id, acknowledged(call, options.ack)));
+    void followAsk(
+      session,
+      thinker,
+      options,
+      captions,
+      phases,
+      running,
+      heard,
+      call,
+      event.blocking,
+      job
+    );
+  }
+}
+async function followAsk(session, thinker, options, captions, phases, running, heard, call, blocking, job) {
+  try {
+    for await (const update of thinker.ask(
+      String(call.arguments.request),
+      heard.slice(-heardLimit),
+      job.abort.signal
+    )) {
+      if (job.abort.signal.aborted) break;
+      if (update.type === "progress") {
+        captions.push({ type: "thinker_progress", callId: call.id, text: update.text });
+        send(() => session.inject(update.text, options.progress));
+        continue;
+      }
+      job.answered = true;
+      captions.push({
+        type: "thinker_result",
+        callId: call.id,
+        summary: textOf(update.message),
+        isError: update.message.isError
+      });
+      if (blocking) send(() => session.inject(textOf(update.message), options.result));
+      else send(() => session.respond(call.id, update.message, options.result));
+    }
+  } catch (error) {
+    if (!job.abort.signal.aborted && !job.answered) {
+      job.answered = true;
+      send(() => session.respond(call.id, errorResult(call, error)));
+    }
+  } finally {
+    running.delete(call.id);
+    phases.push({ phase: "idle" });
+  }
+}
+export {
+  askAgent,
+  bindThinker,
+  talkerTools
+};
+//# sourceMappingURL=talker.js.map

package/dist/talker.js.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../src/talker.ts"],
+  "sourcesContent": ["import {\n  type JsonObject,\n  type Tool,\n  type ToolCall,\n  type ToolResultMessage,\n  Type,\n} from '@earendil-works/pi-ai';\nimport { VoiceRequestError } from './errors.ts';\nimport type { Delivery } from './models.ts';\nimport type { RealtimeSession } from './realtime.ts';\n\nconst heardLimit = 20;\n\nexport const talkerTools = ['ask_agent', 'read'] as const;\n\nexport const askAgent: Tool = {\n  name: 'ask_agent',\n  description:\n    'Hand a request to the coding agent. Use for anything that needs files, the browser, tools or more than a few seconds of thought.',\n  parameters: Type.Object({\n    request: Type.String(),\n    context: Type.Optional(Type.String()),\n  }),\n};\n\nexport type ThinkerUpdate =\n  | { type: 'progress'; text: string }\n  | { type: 'result'; message: ToolResultMessage };\n\nexport interface Thinker {\n  ask(request: string, heard: string[], signal: AbortSignal): AsyncIterable<ThinkerUpdate>;\n}\n\nexport interface BindThinkerOptions {\n  progress: Delivery;\n  result: Delivery;\n  ack: string;\n  directTimeoutMs: number;\n  directMaxChars: number;\n}\n\nexport type ThinkerCaption =\n  | { type: 'thinker_progress'; callId: string; text: string }\n  | { type: 'thinker_result'; callId: string; summary: string; isError: boolean };\n\nexport type ThinkerPhase =\n  | { phase: 'thinking' }\n  | { phase: 'tool'; toolName: string; toolCallId: string; args: JsonObject }\n  | { phase: 'idle' };\n\nexport interface BoundThinker {\n  captions: AsyncIterable<ThinkerCaption>;\n  phases: AsyncIterable<ThinkerPhase>;\n  done: Promise<void>;\n}\n\nclass PushStream<T> implements AsyncIterable<T> {\n  readonly #queue: T[] = [];\n  readonly #waiters: Array<(result: { value?: T; done: boolean }) => void> = [];\n  #closed = false;\n\n  push(value: T): void {\n    if (this.#closed) return;\n    const waiter = this.#waiters.shift();\n    if (waiter) waiter({ value, done: false });\n    else this.#queue.push(value);\n  }\n\n  close(): void {\n    if (this.#closed) return;\n    this.#closed = true;\n    while (this.#waiters.length > 0) this.#waiters.shift()?.({ done: true });\n  }\n\n  async *[Symbol.asyncIterator](): AsyncIterator<T> {\n    while (true) {\n      const queued = this.#queue.shift();\n      if (queued !== undefined) {\n        yield queued;\n        continue;\n      }\n      if (this.#closed) return;\n      const next = await new Promise<{ value?: T; done: boolean }>((resolve) =>\n        this.#waiters.push(resolve)\n      );\n      if (next.done || next.value === undefined) return;\n      yield next.value;\n    }\n  }\n}\n\nfunction send(action: () => void): void {\n  try {\n    action();\n  } catch (error) {\n    if (!(error instanceof VoiceRequestError)) throw error;\n  }\n}\n\nfunction textOf(message: ToolResultMessage): string {\n  return message.content\n    .filter((part) => part.type === 'text')\n    .map((part) => part.text)\n    .join('\\n');\n}\n\nfunction acknowledged(call: ToolCall, ack: string): ToolResultMessage {\n  return {\n    role: 'toolResult',\n    toolCallId: call.id,\n    toolName: call.name,\n    content: [{ type: 'text', text: ack }],\n    isError: false,\n    timestamp: Date.now(),\n  };\n}\n\nfunction errorResult(call: ToolCall, error: unknown): ToolResultMessage {\n  return {\n    role: 'toolResult',\n    toolCallId: call.id,\n    toolName: call.name,\n    content: [{ type: 'text', text: error instanceof Error ? error.message : String(error) }],\n    isError: true,\n    timestamp: Date.now(),\n  };\n}\n\nfunction capResult(result: ToolResultMessage, maxChars: number): ToolResultMessage {\n  let left = maxChars;\n  const content = result.content.map((part) => {\n    if (part.type !== 'text') return part;\n    if (left <= 0) return { ...part, text: '' };\n    if (part.text.length <= left) {\n      left -= part.text.length;\n      return part;\n    }\n    const text = part.text.slice(0, left);\n    left = 0;\n    return { ...part, text };\n  });\n  return { ...result, content };\n}\n\ninterface AskJob {\n  abort: AbortController;\n  answered: boolean;\n  call: ToolCall;\n}\n\nexport function bindThinker(\n  session: RealtimeSession,\n  thinker: Thinker,\n  direct: (call: ToolCall, signal: AbortSignal) => Promise<ToolResultMessage>,\n  options: BindThinkerOptions\n): BoundThinker {\n  const captions = new PushStream<ThinkerCaption>();\n  const phases = new PushStream<ThinkerPhase>();\n  const done = run(session, thinker, direct, options, captions, phases).finally(() => {\n    captions.close();\n    phases.close();\n    captions.close();\n    phases.close();\n  });\n  return { captions, phases, done };\n}\n\nasync function run(\n  session: RealtimeSession,\n  thinker: Thinker,\n  direct: (call: ToolCall, signal: AbortSignal) => Promise<ToolResultMessage>,\n  options: BindThinkerOptions,\n  captions: PushStream<ThinkerCaption>,\n  phases: PushStream<ThinkerPhase>\n): Promise<void> {\n  const heard: string[] = [];\n  const running = new Map<string, AskJob>();\n  for await (const event of session) {\n    if (event.type === 'input_transcript' && event.final) heard.push(`user: ${event.text}`);\n    if (event.type === 'output_transcript' && event.final) heard.push(`talker: ${event.text}`);\n    if (event.type === 'tool_call_cancelled') {\n      const job = running.get(event.callId);\n      if (!job) continue;\n      job.abort.abort();\n      if (!job.answered) {\n        job.answered = true;\n        send(() => session.respond(event.callId, errorResult(job.call, new Error('cancelled'))));\n      }\n      continue;\n    }\n    if (event.type !== 'tool_call') continue;\n    const call = event.call;\n    if (call.name !== askAgent.name) {\n      phases.push({\n        phase: 'tool',\n        toolName: call.name,\n        toolCallId: call.id,\n        args: call.arguments,\n      });\n      void direct(call, AbortSignal.timeout(options.directTimeoutMs))\n        .then((result) => {\n          send(() => session.respond(call.id, capResult(result, options.directMaxChars)));\n          phases.push({ phase: 'idle' });\n        })\n        .catch((error: unknown) => {\n          send(() => session.respond(call.id, errorResult(call, error)));\n          phases.push({ phase: 'idle' });\n        });\n      continue;\n    }\n    const job: AskJob = { abort: new AbortController(), answered: event.blocking, call };\n    running.set(call.id, job);\n    phases.push({ phase: 'thinking' });\n    if (event.blocking) send(() => session.respond(call.id, acknowledged(call, options.ack)));\n    void followAsk(\n      session,\n      thinker,\n      options,\n      captions,\n      phases,\n      running,\n      heard,\n      call,\n      event.blocking,\n      job\n    );\n  }\n}\n\nasync function followAsk(\n  session: RealtimeSession,\n  thinker: Thinker,\n  options: BindThinkerOptions,\n  captions: PushStream<ThinkerCaption>,\n  phases: PushStream<ThinkerPhase>,\n  running: Map<string, AskJob>,\n  heard: string[],\n  call: ToolCall,\n  blocking: boolean,\n  job: AskJob\n): Promise<void> {\n  try {\n    for await (const update of thinker.ask(\n      String(call.arguments.request),\n      heard.slice(-heardLimit),\n      job.abort.signal\n    )) {\n      if (job.abort.signal.aborted) break;\n      if (update.type === 'progress') {\n        captions.push({ type: 'thinker_progress', callId: call.id, text: update.text });\n        send(() => session.inject(update.text, options.progress));\n        continue;\n      }\n      job.answered = true;\n      captions.push({\n        type: 'thinker_result',\n        callId: call.id,\n        summary: textOf(update.message),\n        isError: update.message.isError,\n      });\n      if (blocking) send(() => session.inject(textOf(update.message), options.result));\n      else send(() => session.respond(call.id, update.message, options.result));\n    }\n  } catch (error) {\n    if (!job.abort.signal.aborted && !job.answered) {\n      job.answered = true;\n      send(() => session.respond(call.id, errorResult(call, error)));\n    }\n  } finally {\n    running.delete(call.id);\n    phases.push({ phase: 'idle' });\n  }\n}\n"],
+  "mappings": "AAAA;AAAA,EAKE;AAAA,OACK;AACP,SAAS,yBAAyB;AAIlC,MAAM,aAAa;AAEZ,MAAM,cAAc,CAAC,aAAa,MAAM;AAExC,MAAM,WAAiB;AAAA,EAC5B,MAAM;AAAA,EACN,aACE;AAAA,EACF,YAAY,KAAK,OAAO;AAAA,IACtB,SAAS,KAAK,OAAO;AAAA,IACrB,SAAS,KAAK,SAAS,KAAK,OAAO,CAAC;AAAA,EACtC,CAAC;AACH;AAiCA,MAAM,WAA0C;AAAA,EACrC,SAAc,CAAC;AAAA,EACf,WAAkE,CAAC;AAAA,EAC5E,UAAU;AAAA,EAEV,KAAK,OAAgB;AACnB,QAAI,KAAK,QAAS;AAClB,UAAM,SAAS,KAAK,SAAS,MAAM;AACnC,QAAI,OAAQ,QAAO,EAAE,OAAO,MAAM,MAAM,CAAC;AAAA,QACpC,MAAK,OAAO,KAAK,KAAK;AAAA,EAC7B;AAAA,EAEA,QAAc;AACZ,QAAI,KAAK,QAAS;AAClB,SAAK,UAAU;AACf,WAAO,KAAK,SAAS,SAAS,EAAG,MAAK,SAAS,MAAM,IAAI,EAAE,MAAM,KAAK,CAAC;AAAA,EACzE;AAAA,EAEA,QAAQ,OAAO,aAAa,IAAsB;AAChD,WAAO,MAAM;AACX,YAAM,SAAS,KAAK,OAAO,MAAM;AACjC,UAAI,WAAW,QAAW;AACxB,cAAM;AACN;AAAA,MACF;AACA,UAAI,KAAK,QAAS;AAClB,YAAM,OAAO,MAAM,IAAI;AAAA,QAAsC,CAAC,YAC5D,KAAK,SAAS,KAAK,OAAO;AAAA,MAC5B;AACA,UAAI,KAAK,QAAQ,KAAK,UAAU,OAAW;AAC3C,YAAM,KAAK;AAAA,IACb;AAAA,EACF;AACF;AAEA,SAAS,KAAK,QAA0B;AACtC,MAAI;AACF,WAAO;AAAA,EACT,SAAS,OAAO;AACd,QAAI,EAAE,iBAAiB,mBAAoB,OAAM;AAAA,EACnD;AACF;AAEA,SAAS,OAAO,SAAoC;AAClD,SAAO,QAAQ,QACZ,OAAO,CAAC,SAAS,KAAK,SAAS,MAAM,EACrC,IAAI,CAAC,SAAS,KAAK,IAAI,EACvB,KAAK,IAAI;AACd;AAEA,SAAS,aAAa,MAAgB,KAAgC;AACpE,SAAO;AAAA,IACL,MAAM;AAAA,IACN,YAAY,KAAK;AAAA,IACjB,UAAU,KAAK;AAAA,IACf,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,IAAI,CAAC;AAAA,IACrC,SAAS;AAAA,IACT,WAAW,KAAK,IAAI;AAAA,EACtB;AACF;AAEA,SAAS,YAAY,MAAgB,OAAmC;AACtE,SAAO;AAAA,IACL,MAAM;AAAA,IACN,YAAY,KAAK;AAAA,IACjB,UAAU,KAAK;AAAA,IACf,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK,EAAE,CAAC;AAAA,IACxF,SAAS;AAAA,IACT,WAAW,KAAK,IAAI;AAAA,EACtB;AACF;AAEA,SAAS,UAAU,QAA2B,UAAqC;AACjF,MAAI,OAAO;AACX,QAAM,UAAU,OAAO,QAAQ,IAAI,CAAC,SAAS;AAC3C,QAAI,KAAK,SAAS,OAAQ,QAAO;AACjC,QAAI,QAAQ,EAAG,QAAO,EAAE,GAAG,MAAM,MAAM,GAAG;AAC1C,QAAI,KAAK,KAAK,UAAU,MAAM;AAC5B,cAAQ,KAAK,KAAK;AAClB,aAAO;AAAA,IACT;AACA,UAAM,OAAO,KAAK,KAAK,MAAM,GAAG,IAAI;AACpC,WAAO;AACP,WAAO,EAAE,GAAG,MAAM,KAAK;AAAA,EACzB,CAAC;AACD,SAAO,EAAE,GAAG,QAAQ,QAAQ;AAC9B;AAQO,SAAS,YACd,SACA,SACA,QACA,SACc;AACd,QAAM,WAAW,IAAI,WAA2B;AAChD,QAAM,SAAS,IAAI,WAAyB;AAC5C,QAAM,OAAO,IAAI,SAAS,SAAS,QAAQ,SAAS,UAAU,MAAM,EAAE,QAAQ,MAAM;AAClF,aAAS,MAAM;AACf,WAAO,MAAM;AACb,aAAS,MAAM;AACf,WAAO,MAAM;AAAA,EACf,CAAC;AACD,SAAO,EAAE,UAAU,QAAQ,KAAK;AAClC;AAEA,eAAe,IACb,SACA,SACA,QACA,SACA,UACA,QACe;AACf,QAAM,QAAkB,CAAC;AACzB,QAAM,UAAU,oBAAI,IAAoB;AACxC,mBAAiB,SAAS,SAAS;AACjC,QAAI,MAAM,SAAS,sBAAsB,MAAM,MAAO,OAAM,KAAK,SAAS,MAAM,IAAI,EAAE;AACtF,QAAI,MAAM,SAAS,uBAAuB,MAAM,MAAO,OAAM,KAAK,WAAW,MAAM,IAAI,EAAE;AACzF,QAAI,MAAM,SAAS,uBAAuB;AACxC,YAAMA,OAAM,QAAQ,IAAI,MAAM,MAAM;AACpC,UAAI,CAACA,KAAK;AACV,MAAAA,KAAI,MAAM,MAAM;AAChB,UAAI,CAACA,KAAI,UAAU;AACjB,QAAAA,KAAI,WAAW;AACf,aAAK,MAAM,QAAQ,QAAQ,MAAM,QAAQ,YAAYA,KAAI,MAAM,IAAI,MAAM,WAAW,CAAC,CAAC,CAAC;AAAA,MACzF;AACA;AAAA,IACF;AACA,QAAI,MAAM,SAAS,YAAa;AAChC,UAAM,OAAO,MAAM;AACnB,QAAI,KAAK,SAAS,SAAS,MAAM;AAC/B,aAAO,KAAK;AAAA,QACV,OAAO;AAAA,QACP,UAAU,KAAK;AAAA,QACf,YAAY,KAAK;AAAA,QACjB,MAAM,KAAK;AAAA,MACb,CAAC;AACD,WAAK,OAAO,MAAM,YAAY,QAAQ,QAAQ,eAAe,CAAC,EAC3D,KAAK,CAAC,WAAW;AAChB,aAAK,MAAM,QAAQ,QAAQ,KAAK,IAAI,UAAU,QAAQ,QAAQ,cAAc,CAAC,CAAC;AAC9E,eAAO,KAAK,EAAE,OAAO,OAAO,CAAC;AAAA,MAC/B,CAAC,EACA,MAAM,CAAC,UAAmB;AACzB,aAAK,MAAM,QAAQ,QAAQ,KAAK,IAAI,YAAY,MAAM,KAAK,CAAC,CAAC;AAC7D,eAAO,KAAK,EAAE,OAAO,OAAO,CAAC;AAAA,MAC/B,CAAC;AACH;AAAA,IACF;AACA,UAAM,MAAc,EAAE,OAAO,IAAI,gBAAgB,GAAG,UAAU,MAAM,UAAU,KAAK;AACnF,YAAQ,IAAI,KAAK,IAAI,GAAG;AACxB,WAAO,KAAK,EAAE,OAAO,WAAW,CAAC;AACjC,QAAI,MAAM,SAAU,MAAK,MAAM,QAAQ,QAAQ,KAAK,IAAI,aAAa,MAAM,QAAQ,GAAG,CAAC,CAAC;AACxF,SAAK;AAAA,MACH;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,MAAM;AAAA,MACN;AAAA,IACF;AAAA,EACF;AACF;AAEA,eAAe,UACb,SACA,SACA,SACA,UACA,QACA,SACA,OACA,MACA,UACA,KACe;AACf,MAAI;AACF,qBAAiB,UAAU,QAAQ;AAAA,MACjC,OAAO,KAAK,UAAU,OAAO;AAAA,MAC7B,MAAM,MAAM,CAAC,UAAU;AAAA,MACvB,IAAI,MAAM;AAAA,IACZ,GAAG;AACD,UAAI,IAAI,MAAM,OAAO,QAAS;AAC9B,UAAI,OAAO,SAAS,YAAY;AAC9B,iBAAS,KAAK,EAAE,MAAM,oBAAoB,QAAQ,KAAK,IAAI,MAAM,OAAO,KAAK,CAAC;AAC9E,aAAK,MAAM,QAAQ,OAAO,OAAO,MAAM,QAAQ,QAAQ,CAAC;AACxD;AAAA,MACF;AACA,UAAI,WAAW;AACf,eAAS,KAAK;AAAA,QACZ,MAAM;AAAA,QACN,QAAQ,KAAK;AAAA,QACb,SAAS,OAAO,OAAO,OAAO;AAAA,QAC9B,SAAS,OAAO,QAAQ;AAAA,MAC1B,CAAC;AACD,UAAI,SAAU,MAAK,MAAM,QAAQ,OAAO,OAAO,OAAO,OAAO,GAAG,QAAQ,MAAM,CAAC;AAAA,UAC1E,MAAK,MAAM,QAAQ,QAAQ,KAAK,IAAI,OAAO,SAAS,QAAQ,MAAM,CAAC;AAAA,IAC1E;AAAA,EACF,SAAS,OAAO;AACd,QAAI,CAAC,IAAI,MAAM,OAAO,WAAW,CAAC,IAAI,UAAU;AAC9C,UAAI,WAAW;AACf,WAAK,MAAM,QAAQ,QAAQ,KAAK,IAAI,YAAY,MAAM,KAAK,CAAC,CAAC;AAAA,IAC/D;AAAA,EACF,UAAE;AACA,YAAQ,OAAO,KAAK,EAAE;AACtB,WAAO,KAAK,EAAE,OAAO,OAAO,CAAC;AAAA,EAC/B;AACF;",
+  "names": ["job"]
+}

package/dist/types/index.d.ts

@@ -10,6 +10,8 @@ export { planDelivery, planTruncation, RealtimeLedger, realtimeSocket } from './
 export { connectRealtime, listen, registerRealtimeApiProvider, registerSpeechApiProvider, registerTranscriptionApiProvider, speak, transcribe, } from './registry.ts';
 export type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';
 export { SpeechEventStream } from './speech.ts';
+export type { BindThinkerOptions, BoundThinker, Thinker, ThinkerCaption, ThinkerPhase, ThinkerUpdate, } from './talker.ts';
+export { askAgent, bindThinker, talkerTools } from './talker.ts';
 export type { AudioInput, ListenOptions, Transcript, TranscriptionApiProvider, TranscriptionEvent, TranscriptSegment, } from './transcription.ts';
 export { TranscriptionEventStream } from './transcription.ts';
 export type { AudioCost, AudioCounters, AudioUsage, RealtimeAudioCost, SpeechAudioCost, TranscriptionAudioCost, } from './usage.ts';

package/dist/types/talker.d.ts

@@ -0,0 +1,48 @@
+import { type JsonObject, type Tool, type ToolCall, type ToolResultMessage } from '@earendil-works/pi-ai';
+import type { Delivery } from './models.ts';
+import type { RealtimeSession } from './realtime.ts';
+export declare const talkerTools: readonly ['ask_agent', 'read'];
+export declare const askAgent: Tool;
+export type ThinkerUpdate = {
+    type: 'progress';
+    text: string;
+} | {
+    type: 'result';
+    message: ToolResultMessage;
+};
+export interface Thinker {
+    ask(request: string, heard: string[], signal: AbortSignal): AsyncIterable<ThinkerUpdate>;
+}
+export interface BindThinkerOptions {
+    progress: Delivery;
+    result: Delivery;
+    ack: string;
+    directTimeoutMs: number;
+    directMaxChars: number;
+}
+export type ThinkerCaption = {
+    type: 'thinker_progress';
+    callId: string;
+    text: string;
+} | {
+    type: 'thinker_result';
+    callId: string;
+    summary: string;
+    isError: boolean;
+};
+export type ThinkerPhase = {
+    phase: 'thinking';
+} | {
+    phase: 'tool';
+    toolName: string;
+    toolCallId: string;
+    args: JsonObject;
+} | {
+    phase: 'idle';
+};
+export interface BoundThinker {
+    captions: AsyncIterable<ThinkerCaption>;
+    phases: AsyncIterable<ThinkerPhase>;
+    done: Promise<void>;
+}
+export declare function bindThinker(session: RealtimeSession, thinker: Thinker, direct: (call: ToolCall, signal: AbortSignal) => Promise<ToolResultMessage>, options: BindThinkerOptions): BoundThinker;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-ecoverse/pi-audio",
-  "version": "1.6.0",
+  "version": "1.7.0",
   "description": "Voice bindings for pi: speech-to-text, text-to-speech and realtime speech for pi-coding-agent plugins and slicc-agent",
   "type": "module",
   "license": "Apache-2.0",