@vibesdotdev/agent-execution 0.0.0-stage → 1.0.0-rc.2

package/dist/voice/agent-voice.descriptor.d.ts

@@ -0,0 +1,82 @@
+/**
+ * agent/voice descriptor schema + factory
+ *
+ * Defines the `agent/voice` runtime kind: a half-duplex (turn-based) voice
+ * pipeline that wires three existing seams together —
+ *
+ *   STT (ai/generation transcription) → text → NativeToolsAgentLoop (text-only)
+ *   → text → TTS (ai/generation tts)
+ *
+ * The agent loop is left completely unchanged: the voice kind is an
+ * orchestrator that converts speech↔text at the edges and delegates the
+ * "what should the assistant say" decision to the loop. `interruptible` is
+ * fixed to `false` this round — barge-in/full-duplex is a follow-up.
+ *
+ * Provider ids reference `ai/generation` providers (STT:
+ * deepgram/openai-transcription/whisper-cpp/faster-whisper; TTS:
+ * elevenlabs/openai-tts/bark/coqui-tts). The kind does NOT call any cloud AI
+ * directly — it resolves providers through the runtime's generation seam.
+ *
+ * This file is metadata-only (zod + runtime descriptor schema) and safe to
+ * import anywhere, including workerd.
+ */
+import * as z from 'zod/v4';
+export declare const VOICE_KIND_ID: "agent/voice";
+/** PCM/container formats the pipeline accepts on the wire. Mirrors the voice-conversation recipe. */
+export declare const VOICE_AUDIO_FORMATS: readonly ["pcm", "wav", "mp3", "opus", "aac", "flac", "webm"];
+export declare const VoiceAudioFormatSchema: z.ZodEnum<{
+    pcm: "pcm";
+    wav: "wav";
+    mp3: "mp3";
+    opus: "opus";
+    aac: "aac";
+    flac: "flac";
+    webm: "webm";
+}>;
+export type VoiceAudioFormat = z.infer<typeof VoiceAudioFormatSchema>;
+export declare const AgentVoiceDescriptorSchema: z.ZodObject<{
+    id: z.ZodString;
+    name: z.ZodOptional<z.ZodString>;
+    description: z.ZodOptional<z.ZodString>;
+    tags: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    hardware: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    enabled: z.ZodOptional<z.ZodBoolean>;
+    capabilityId: z.ZodOptional<z.ZodString>;
+    requiredContext: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    optionalContext: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    config: z.ZodOptional<z.ZodUnknown>;
+    retention: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
+        maxAgeDays: z.ZodOptional<z.ZodNumber>;
+        maxEntries: z.ZodOptional<z.ZodNumber>;
+        maxBytes: z.ZodOptional<z.ZodNumber>;
+        cascadeFrom: z.ZodOptional<z.ZodString>;
+        ageField: z.ZodOptional<z.ZodString>;
+        reason: z.ZodString;
+        enforcedBy: z.ZodOptional<z.ZodString>;
+    }, z.core.$strip>]>>;
+    kind: z.ZodLiteral<"agent/voice">;
+    sttProviderId: z.ZodString;
+    sttModelId: z.ZodOptional<z.ZodString>;
+    ttsProviderId: z.ZodString;
+    ttsModelId: z.ZodOptional<z.ZodString>;
+    agentLoopId: z.ZodDefault<z.ZodString>;
+    sampleRate: z.ZodDefault<z.ZodNumber>;
+    audioFormat: z.ZodDefault<z.ZodEnum<{
+        pcm: "pcm";
+        wav: "wav";
+        mp3: "mp3";
+        opus: "opus";
+        aac: "aac";
+        flac: "flac";
+        webm: "webm";
+    }>>;
+    turnTimeoutMs: z.ZodDefault<z.ZodNumber>;
+    silenceTimeoutMs: z.ZodDefault<z.ZodNumber>;
+    silenceThreshold: z.ZodDefault<z.ZodNumber>;
+    interruptible: z.ZodDefault<z.ZodLiteral<false>>;
+    priority: z.ZodDefault<z.ZodNumber>;
+}, z.core.$strip>;
+export type AgentVoiceDescriptor = z.infer<typeof AgentVoiceDescriptorSchema>;
+/** Author an `agent/voice` descriptor without restating the kind id. */
+export declare function createAgentVoiceDescriptor(options: Omit<z.input<typeof AgentVoiceDescriptorSchema>, 'kind'>): AgentVoiceDescriptor;
+//# sourceMappingURL=agent-voice.descriptor.d.ts.map

package/dist/voice/agent-voice.descriptor.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"agent-voice.descriptor.d.ts","sourceRoot":"","sources":["../../src/voice/agent-voice.descriptor.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;GAqBG;AAEH,OAAO,KAAK,CAAC,MAAM,QAAQ,CAAC;AAG5B,eAAO,MAAM,aAAa,EAAG,aAAsB,CAAC;AAEpD,qGAAqG;AACrG,eAAO,MAAM,mBAAmB,+DAAgE,CAAC;AACjG,eAAO,MAAM,sBAAsB;;;;;;;;EAA8B,CAAC;AAClE,MAAM,MAAM,gBAAgB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,sBAAsB,CAAC,CAAC;AAEtE,eAAO,MAAM,0BAA0B;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;iBAoDrC,CAAC;AAEH,MAAM,MAAM,oBAAoB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,0BAA0B,CAAC,CAAC;AAE9E,wEAAwE;AACxE,wBAAgB,0BAA0B,CACzC,OAAO,EAAE,IAAI,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,0BAA0B,CAAC,EAAE,MAAM,CAAC,GAC/D,oBAAoB,CAKtB"}

package/dist/voice/agent-voice.descriptor.js

@@ -0,0 +1,79 @@
+/**
+ * agent/voice descriptor schema + factory
+ *
+ * Defines the `agent/voice` runtime kind: a half-duplex (turn-based) voice
+ * pipeline that wires three existing seams together —
+ *
+ *   STT (ai/generation transcription) → text → NativeToolsAgentLoop (text-only)
+ *   → text → TTS (ai/generation tts)
+ *
+ * The agent loop is left completely unchanged: the voice kind is an
+ * orchestrator that converts speech↔text at the edges and delegates the
+ * "what should the assistant say" decision to the loop. `interruptible` is
+ * fixed to `false` this round — barge-in/full-duplex is a follow-up.
+ *
+ * Provider ids reference `ai/generation` providers (STT:
+ * deepgram/openai-transcription/whisper-cpp/faster-whisper; TTS:
+ * elevenlabs/openai-tts/bark/coqui-tts). The kind does NOT call any cloud AI
+ * directly — it resolves providers through the runtime's generation seam.
+ *
+ * This file is metadata-only (zod + runtime descriptor schema) and safe to
+ * import anywhere, including workerd.
+ */
+import * as z from 'zod/v4';
+import { RuntimeDescriptorSchema } from '@vibesdotdev/kernel/schemas/descriptor';
+export const VOICE_KIND_ID = 'agent/voice';
+/** PCM/container formats the pipeline accepts on the wire. Mirrors the voice-conversation recipe. */
+export const VOICE_AUDIO_FORMATS = ['pcm', 'wav', 'mp3', 'opus', 'aac', 'flac', 'webm'];
+export const VoiceAudioFormatSchema = z.enum(VOICE_AUDIO_FORMATS);
+export const AgentVoiceDescriptorSchema = RuntimeDescriptorSchema.extend({
+    kind: z.literal(VOICE_KIND_ID),
+    /** ai/generation provider id used for speech→text. */
+    sttProviderId: z.string().min(1),
+    /** Transcription model id for the STT provider. */
+    sttModelId: z.string().min(1).optional(),
+    /** ai/generation provider id used for text→speech. */
+    ttsProviderId: z.string().min(1),
+    /** TTS model id for the TTS provider. */
+    ttsModelId: z.string().min(1).optional(),
+    /**
+     * Agent loop id that drives the text turn. Defaults to the canonical
+     * native function loop descriptor. The loop runs text-only and is never
+     * modified by the voice kind.
+     */
+    agentLoopId: z.string().min(1).default('native-function-loop'),
+    /** Sample rate of the inbound/outbound PCM stream (Hz). */
+    sampleRate: z.number().int().positive().default(16_000),
+    /** Wire audio format for the turn frames. */
+    audioFormat: VoiceAudioFormatSchema.default('pcm'),
+    /**
+     * How long the server waits for a turn to complete (silence-terminated
+     * utterance → STT → agent → TTS) before timing out. Half-duplex: one turn
+     * at a time.
+     */
+    turnTimeoutMs: z.number().int().positive().default(60_000),
+    /**
+     * Trailing silence (ms) that terminates an utterance. Server-side silence
+     * detection segments the inbound stream into turns.
+     */
+    silenceTimeoutMs: z.number().int().positive().default(1_200),
+    /**
+     * Amplitude below which a PCM frame counts as silence (0..1 normalized
+     * RMS). Tuned conservatively for 16-bit PCM.
+     */
+    silenceThreshold: z.number().min(0).max(1).default(0.02),
+    /**
+     * Fixed `false` this round — half-duplex only. A future full-duplex impl
+     * would set this true and add barge-in.
+     */
+    interruptible: z.literal(false).default(false),
+    priority: z.number().int().default(100)
+});
+/** Author an `agent/voice` descriptor without restating the kind id. */
+export function createAgentVoiceDescriptor(options) {
+    return AgentVoiceDescriptorSchema.parse({
+        ...options,
+        kind: VOICE_KIND_ID
+    });
+}
+//# sourceMappingURL=agent-voice.descriptor.js.map

package/dist/voice/agent-voice.descriptor.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"agent-voice.descriptor.js","sourceRoot":"","sources":["../../src/voice/agent-voice.descriptor.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;GAqBG;AAEH,OAAO,KAAK,CAAC,MAAM,QAAQ,CAAC;AAC5B,OAAO,EAAE,uBAAuB,EAAE,MAAM,wCAAwC,CAAC;AAEjF,MAAM,CAAC,MAAM,aAAa,GAAG,aAAsB,CAAC;AAEpD,qGAAqG;AACrG,MAAM,CAAC,MAAM,mBAAmB,GAAG,CAAC,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,EAAE,MAAM,CAAU,CAAC;AACjG,MAAM,CAAC,MAAM,sBAAsB,GAAG,CAAC,CAAC,IAAI,CAAC,mBAAmB,CAAC,CAAC;AAGlE,MAAM,CAAC,MAAM,0BAA0B,GAAG,uBAAuB,CAAC,MAAM,CAAC;IACxE,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,aAAa,CAAC;IAE9B,sDAAsD;IACtD,aAAa,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC;IAChC,mDAAmD;IACnD,UAAU,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,EAAE;IAExC,sDAAsD;IACtD,aAAa,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC;IAChC,yCAAyC;IACzC,UAAU,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,EAAE;IAExC;;;;OAIG;IACH,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,sBAAsB,CAAC;IAE9D,2DAA2D;IAC3D,UAAU,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,MAAM,CAAC;IAEvD,6CAA6C;IAC7C,WAAW,EAAE,sBAAsB,CAAC,OAAO,CAAC,KAAK,CAAC;IAElD;;;;OAIG;IACH,aAAa,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,MAAM,CAAC;IAE1D;;;OAGG;IACH,gBAAgB,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,KAAK,CAAC;IAE5D;;;OAGG;IACH,gBAAgB,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,IAAI,CAAC;IAExD;;;OAGG;IACH,aAAa,EAAE,CAAC,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC,OAAO,CAAC,KAAK,CAAC;IAE9C,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,OAAO,CAAC,GAAG,CAAC;CACvC,CAAC,CAAC;AAIH,wEAAwE;AACxE,MAAM,UAAU,0BAA0B,CACzC,OAAiE;IAEjE,OAAO,0BAA0B,CAAC,KAAK,CAAC;QACvC,GAAG,OAAO;QACV,IAAI,EAAE,aAAa;KACnB,CAAC,CAAC;AACJ,CAAC"}

package/dist/voice/agent-voice.impl.consumer.d.ts

@@ -0,0 +1,35 @@
+/**
+ * agent/voice — local (consumer) implementation.
+ *
+ * `hardware:['consumer']`. Holds the validated descriptor and exposes a
+ * half-duplex *session* that the transport drives:
+ *
+ *   - `pushFrame(pcm)`  — feed an inbound PCM frame; server-side silence
+ *     detection segments turns. When an utterance completes, the session runs
+ *     STT → agent loop → TTS via the injected {@link VoicePipelineDeps} and
+ *     returns the turn result (transcript + reply text + synthesized audio).
+ *   - `submitUtterance(audio)` — for transports that pre-segment (e.g. the
+ *     POST turn endpoint), skip silence detection and run the turn directly.
+ *
+ * Conversation history is in-memory per session (MVP). The agent loop itself
+ * is untouched — the session only converts speech↔text at the edges.
+ *
+ * workerd-safe: depends on the runtime + ai schemas + the dependency-injected
+ * pipeline. No Node-only imports.
+ */
+import type { KindContext } from '@vibesdotdev/kernel/schemas/kind';
+import type { RuntimeDescriptor } from '@vibesdotdev/kernel/schemas/descriptor';
+import { type AgentVoiceDescriptor } from './agent-voice.descriptor.ts';
+import type { VoicePipelineDeps } from './pipeline.ts';
+import { VoiceSession } from './session.ts';
+export { VoiceSession } from './session.ts';
+export declare class ConsumerAgentVoice {
+    readonly descriptor: AgentVoiceDescriptor;
+    /** The resolving runtime owns logging, clocks, cancellation and every turn's bounded wait. */
+    private readonly runtime;
+    constructor(descriptor: RuntimeDescriptor | AgentVoiceDescriptor, context: KindContext);
+    /** Open a half-duplex session bound to the injected STT/agent/TTS stages. */
+    createSession(deps: VoicePipelineDeps, systemPrompt?: string): VoiceSession;
+}
+export default ConsumerAgentVoice;
+//# sourceMappingURL=agent-voice.impl.consumer.d.ts.map

package/dist/voice/agent-voice.impl.consumer.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"agent-voice.impl.consumer.d.ts","sourceRoot":"","sources":["../../src/voice/agent-voice.impl.consumer.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;GAkBG;AAGH,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,kCAAkC,CAAC;AACpE,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,wCAAwC,CAAC;AAChF,OAAO,EAEN,KAAK,oBAAoB,EACzB,MAAM,6BAA6B,CAAC;AAIrC,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,eAAe,CAAC;AACvD,OAAO,EAAE,YAAY,EAAE,MAAM,cAAc,CAAC;AAC5C,OAAO,EAAE,YAAY,EAAE,MAAM,cAAc,CAAC;AAE5C,qBAAa,kBAAkB;IAC9B,QAAQ,CAAC,UAAU,EAAE,oBAAoB,CAAC;IAC1C,8FAA8F;IAC9F,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAa;gBAEzB,UAAU,EAAE,iBAAiB,GAAG,oBAAoB,EAAE,OAAO,EAAE,WAAW;IAStF,6EAA6E;IAC7E,aAAa,CAAC,IAAI,EAAE,iBAAiB,EAAE,YAAY,CAAC,EAAE,MAAM,GAAG,YAAY;CAO3E;AAED,eAAe,kBAAkB,CAAC"}

package/dist/voice/agent-voice.impl.consumer.js

@@ -0,0 +1,50 @@
+/**
+ * agent/voice — local (consumer) implementation.
+ *
+ * `hardware:['consumer']`. Holds the validated descriptor and exposes a
+ * half-duplex *session* that the transport drives:
+ *
+ *   - `pushFrame(pcm)`  — feed an inbound PCM frame; server-side silence
+ *     detection segments turns. When an utterance completes, the session runs
+ *     STT → agent loop → TTS via the injected {@link VoicePipelineDeps} and
+ *     returns the turn result (transcript + reply text + synthesized audio).
+ *   - `submitUtterance(audio)` — for transports that pre-segment (e.g. the
+ *     POST turn endpoint), skip silence detection and run the turn directly.
+ *
+ * Conversation history is in-memory per session (MVP). The agent loop itself
+ * is untouched — the session only converts speech↔text at the edges.
+ *
+ * workerd-safe: depends on the runtime + ai schemas + the dependency-injected
+ * pipeline. No Node-only imports.
+ */
+import {} from '@vibesdotdev/kernel';
+import { AgentVoiceDescriptorSchema } from "./agent-voice.descriptor.js";
+import { tryGetVibesClient } from '@vibesdotdev/kernel/client';
+import { CapabilityError } from '@vibesdotdev/kernel';
+import { VoiceClient, voiceClientDescriptor } from "./client.js";
+import { VoiceSession } from "./session.js";
+export { VoiceSession } from "./session.js";
+export class ConsumerAgentVoice {
+    descriptor;
+    /** The resolving runtime owns logging, clocks, cancellation and every turn's bounded wait. */
+    runtime;
+    constructor(descriptor, context) {
+        // Defensive parse so a hand-authored descriptor still validates here even
+        // if it skipped the factory. Constructor shape `(descriptor, context)`
+        // matches the kind's default-implementation contract so the runtime can
+        // build this either via the loader or the kind default.
+        this.descriptor = AgentVoiceDescriptorSchema.parse(descriptor);
+        this.runtime = context.runtime;
+    }
+    /** Open a half-duplex session bound to the injected STT/agent/TTS stages. */
+    createSession(deps, systemPrompt) {
+        const client = tryGetVibesClient(voiceClientDescriptor.id, this.runtime);
+        if (!client)
+            throw new CapabilityError('The voice runtime is unavailable.', 'UNAVAILABLE', {
+                reason: 'voice-session-closed', retryable: false
+            });
+        return client.createSession(this.descriptor, deps, systemPrompt);
+    }
+}
+export default ConsumerAgentVoice;
+//# sourceMappingURL=agent-voice.impl.consumer.js.map

package/dist/voice/agent-voice.impl.consumer.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"agent-voice.impl.consumer.js","sourceRoot":"","sources":["../../src/voice/agent-voice.impl.consumer.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;GAkBG;AAEH,OAAO,EAAmB,MAAM,qBAAqB,CAAC;AAGtD,OAAO,EACN,0BAA0B,EAE1B,MAAM,6BAA6B,CAAC;AACrC,OAAO,EAAE,iBAAiB,EAAE,MAAM,4BAA4B,CAAC;AAC/D,OAAO,EAAE,eAAe,EAAE,MAAM,qBAAqB,CAAC;AACtD,OAAO,EAAE,WAAW,EAAE,qBAAqB,EAAE,MAAM,aAAa,CAAC;AAEjE,OAAO,EAAE,YAAY,EAAE,MAAM,cAAc,CAAC;AAC5C,OAAO,EAAE,YAAY,EAAE,MAAM,cAAc,CAAC;AAE5C,MAAM,OAAO,kBAAkB;IACrB,UAAU,CAAuB;IAC1C,8FAA8F;IAC7E,OAAO,CAAa;IAErC,YAAY,UAAoD,EAAE,OAAoB;QACrF,0EAA0E;QAC1E,uEAAuE;QACvE,wEAAwE;QACxE,wDAAwD;QACxD,IAAI,CAAC,UAAU,GAAG,0BAA0B,CAAC,KAAK,CAAC,UAAU,CAAC,CAAC;QAC/D,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC,OAAO,CAAC;IAChC,CAAC;IAED,6EAA6E;IAC7E,aAAa,CAAC,IAAuB,EAAE,YAAqB;QAC3D,MAAM,MAAM,GAAG,iBAAiB,CAAc,qBAAqB,CAAC,EAAE,EAAE,IAAI,CAAC,OAAO,CAAC,CAAC;QACtF,IAAI,CAAC,MAAM;YAAE,MAAM,IAAI,eAAe,CAAC,mCAAmC,EAAE,aAAa,EAAE;gBAC1F,MAAM,EAAE,sBAAsB,EAAE,SAAS,EAAE,KAAK;aAChD,CAAC,CAAC;QACH,OAAO,MAAM,CAAC,aAAa,CAAC,IAAI,CAAC,UAAU,EAAE,IAAI,EAAE,YAAY,CAAC,CAAC;IAClE,CAAC;CACD;AAED,eAAe,kBAAkB,CAAC"}

package/dist/voice/agent-voice.kind.d.ts

@@ -0,0 +1,67 @@
+/**
+ * `agent/voice` runtime kind descriptor record.
+ *
+ * Plain `RuntimeKindDescriptorRecord` registered via the agent-voice plugin's
+ * `kinds:[...]` — mirrors `agents/src/runtime-kinds.ts`. The kind has
+ * a `defaultImplementation` that simply parses + holds the descriptor (the
+ * live behavior is in the consumer impl, registered as a loader with
+ * `hardware:['consumer']`). Resolution is the runtime default (hardware +
+ * priority); no custom resolver is needed.
+ */
+import type { KindContext } from '@vibesdotdev/kernel/schemas/kind';
+import type { RuntimeDescriptor } from '@vibesdotdev/kernel/schemas/descriptor';
+import { type AgentVoiceDescriptor } from './agent-voice.descriptor.ts';
+declare class DefaultAgentVoice {
+    readonly descriptor: AgentVoiceDescriptor;
+    constructor(descriptor: RuntimeDescriptor, _context: KindContext);
+}
+export declare const agentVoiceKind: {
+    id: "agent/voice";
+    descriptorSchema: import("zod/v4").ZodObject<{
+        id: import("zod/v4").ZodString;
+        name: import("zod/v4").ZodOptional<import("zod/v4").ZodString>;
+        description: import("zod/v4").ZodOptional<import("zod/v4").ZodString>;
+        tags: import("zod/v4").ZodOptional<import("zod/v4").ZodArray<import("zod/v4").ZodString>>;
+        hardware: import("zod/v4").ZodOptional<import("zod/v4").ZodArray<import("zod/v4").ZodString>>;
+        enabled: import("zod/v4").ZodOptional<import("zod/v4").ZodBoolean>;
+        capabilityId: import("zod/v4").ZodOptional<import("zod/v4").ZodString>;
+        requiredContext: import("zod/v4").ZodOptional<import("zod/v4").ZodArray<import("zod/v4").ZodString>>;
+        optionalContext: import("zod/v4").ZodOptional<import("zod/v4").ZodArray<import("zod/v4").ZodString>>;
+        config: import("zod/v4").ZodOptional<import("zod/v4").ZodUnknown>;
+        retention: import("zod/v4").ZodOptional<import("zod/v4").ZodUnion<readonly [import("zod/v4").ZodString, import("zod/v4").ZodObject<{
+            maxAgeDays: import("zod/v4").ZodOptional<import("zod/v4").ZodNumber>;
+            maxEntries: import("zod/v4").ZodOptional<import("zod/v4").ZodNumber>;
+            maxBytes: import("zod/v4").ZodOptional<import("zod/v4").ZodNumber>;
+            cascadeFrom: import("zod/v4").ZodOptional<import("zod/v4").ZodString>;
+            ageField: import("zod/v4").ZodOptional<import("zod/v4").ZodString>;
+            reason: import("zod/v4").ZodString;
+            enforcedBy: import("zod/v4").ZodOptional<import("zod/v4").ZodString>;
+        }, import("zod/v4/core").$strip>]>>;
+        kind: import("zod/v4").ZodLiteral<"agent/voice">;
+        sttProviderId: import("zod/v4").ZodString;
+        sttModelId: import("zod/v4").ZodOptional<import("zod/v4").ZodString>;
+        ttsProviderId: import("zod/v4").ZodString;
+        ttsModelId: import("zod/v4").ZodOptional<import("zod/v4").ZodString>;
+        agentLoopId: import("zod/v4").ZodDefault<import("zod/v4").ZodString>;
+        sampleRate: import("zod/v4").ZodDefault<import("zod/v4").ZodNumber>;
+        audioFormat: import("zod/v4").ZodDefault<import("zod/v4").ZodEnum<{
+            pcm: "pcm";
+            wav: "wav";
+            mp3: "mp3";
+            opus: "opus";
+            aac: "aac";
+            flac: "flac";
+            webm: "webm";
+        }>>;
+        turnTimeoutMs: import("zod/v4").ZodDefault<import("zod/v4").ZodNumber>;
+        silenceTimeoutMs: import("zod/v4").ZodDefault<import("zod/v4").ZodNumber>;
+        silenceThreshold: import("zod/v4").ZodDefault<import("zod/v4").ZodNumber>;
+        interruptible: import("zod/v4").ZodDefault<import("zod/v4").ZodLiteral<false>>;
+        priority: import("zod/v4").ZodDefault<import("zod/v4").ZodNumber>;
+    }, import("zod/v4/core").$strip>;
+    defaultImplementation: typeof DefaultAgentVoice;
+    resolve: (candidates: RuntimeDescriptor[]) => RuntimeDescriptor | undefined;
+    contexts: never[];
+};
+export default agentVoiceKind;
+//# sourceMappingURL=agent-voice.kind.d.ts.map

package/dist/voice/agent-voice.kind.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"agent-voice.kind.d.ts","sourceRoot":"","sources":["../../src/voice/agent-voice.kind.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAEH,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,kCAAkC,CAAC;AACpE,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,wCAAwC,CAAC;AAChF,OAAO,EAGN,KAAK,oBAAoB,EACzB,MAAM,6BAA6B,CAAC;AAErC,cAAM,iBAAiB;IACtB,QAAQ,CAAC,UAAU,EAAE,oBAAoB,CAAC;gBAC9B,UAAU,EAAE,iBAAiB,EAAE,QAAQ,EAAE,WAAW;CAGhE;AAED,eAAO,MAAM,cAAc;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;0BAIJ,iBAAiB,EAAE,KAAG,iBAAiB,GAAG,SAAS;;CAEzE,CAAC;AAEF,eAAe,cAAc,CAAC"}

package/dist/voice/agent-voice.kind.js

@@ -0,0 +1,26 @@
+/**
+ * `agent/voice` runtime kind descriptor record.
+ *
+ * Plain `RuntimeKindDescriptorRecord` registered via the agent-voice plugin's
+ * `kinds:[...]` — mirrors `agents/src/runtime-kinds.ts`. The kind has
+ * a `defaultImplementation` that simply parses + holds the descriptor (the
+ * live behavior is in the consumer impl, registered as a loader with
+ * `hardware:['consumer']`). Resolution is the runtime default (hardware +
+ * priority); no custom resolver is needed.
+ */
+import { AgentVoiceDescriptorSchema, VOICE_KIND_ID } from "./agent-voice.descriptor.js";
+class DefaultAgentVoice {
+    descriptor;
+    constructor(descriptor, _context) {
+        this.descriptor = AgentVoiceDescriptorSchema.parse(descriptor);
+    }
+}
+export const agentVoiceKind = {
+    id: VOICE_KIND_ID,
+    descriptorSchema: AgentVoiceDescriptorSchema,
+    defaultImplementation: DefaultAgentVoice,
+    resolve: (candidates) => candidates[0],
+    contexts: []
+};
+export default agentVoiceKind;
+//# sourceMappingURL=agent-voice.kind.js.map

package/dist/voice/agent-voice.kind.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"agent-voice.kind.js","sourceRoot":"","sources":["../../src/voice/agent-voice.kind.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAIH,OAAO,EACN,0BAA0B,EAC1B,aAAa,EAEb,MAAM,6BAA6B,CAAC;AAErC,MAAM,iBAAiB;IACb,UAAU,CAAuB;IAC1C,YAAY,UAA6B,EAAE,QAAqB;QAC/D,IAAI,CAAC,UAAU,GAAG,0BAA0B,CAAC,KAAK,CAAC,UAAU,CAAC,CAAC;IAChE,CAAC;CACD;AAED,MAAM,CAAC,MAAM,cAAc,GAAG;IAC7B,EAAE,EAAE,aAAa;IACjB,gBAAgB,EAAE,0BAA0B;IAC5C,qBAAqB,EAAE,iBAAiB;IACxC,OAAO,EAAE,CAAC,UAA+B,EAAiC,EAAE,CAAC,UAAU,CAAC,CAAC,CAAC;IAC1F,QAAQ,EAAE,EAAE;CACZ,CAAC;AAEF,eAAe,cAAc,CAAC"}

package/dist/voice/agent-voice.plugin.d.ts

@@ -0,0 +1,4 @@
+/** Universal voice kind and semantic client; consumer loaders have their own slice. */
+export declare const agentVoicePlugin: import("@vibesdotdev/kernel").RuntimePlugin;
+export default agentVoicePlugin;
+//# sourceMappingURL=agent-voice.plugin.d.ts.map

package/dist/voice/agent-voice.plugin.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"agent-voice.plugin.d.ts","sourceRoot":"","sources":["../../src/voice/agent-voice.plugin.ts"],"names":[],"mappings":"AAKA,uFAAuF;AACvF,eAAO,MAAM,gBAAgB,6CAU3B,CAAC;AAEH,eAAe,gBAAgB,CAAC"}

package/dist/voice/agent-voice.plugin.js

@@ -0,0 +1,18 @@
+import { createRuntimePlugin, loader } from '@vibesdotdev/kernel';
+import { getVibesClient } from '@vibesdotdev/kernel/client';
+import { agentVoiceKind } from "./agent-voice.kind.js";
+import { voiceClientDescriptor } from "./client.js";
+/** Universal voice kind and semantic client; consumer loaders have their own slice. */
+export const agentVoicePlugin = createRuntimePlugin({
+    id: 'agent-voice',
+    name: 'Agent Voice',
+    dependencies: ['runtime-client'],
+    kinds: [agentVoiceKind],
+    descriptors: [voiceClientDescriptor],
+    loaders: [loader(voiceClientDescriptor.id, { kind: 'runtime/client' }, () => import("./client.js"))],
+    async onRegister(runtime) {
+        await getVibesClient(voiceClientDescriptor.id, runtime);
+    }
+});
+export default agentVoicePlugin;
+//# sourceMappingURL=agent-voice.plugin.js.map

package/dist/voice/agent-voice.plugin.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"agent-voice.plugin.js","sourceRoot":"","sources":["../../src/voice/agent-voice.plugin.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,mBAAmB,EAAE,MAAM,EAAE,MAAM,qBAAqB,CAAC;AAClE,OAAO,EAAE,cAAc,EAAE,MAAM,4BAA4B,CAAC;AAC5D,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AACvD,OAAO,EAAE,qBAAqB,EAAoB,MAAM,aAAa,CAAC;AAEtE,uFAAuF;AACvF,MAAM,CAAC,MAAM,gBAAgB,GAAG,mBAAmB,CAAC;IACnD,EAAE,EAAE,aAAa;IACjB,IAAI,EAAE,aAAa;IACnB,YAAY,EAAE,CAAC,gBAAgB,CAAC;IAChC,KAAK,EAAE,CAAC,cAAc,CAAC;IACvB,WAAW,EAAE,CAAC,qBAAqB,CAAC;IACpC,OAAO,EAAE,CAAC,MAAM,CAAC,qBAAqB,CAAC,EAAE,EAAE,EAAE,IAAI,EAAE,gBAAgB,EAAE,EAAE,GAAG,EAAE,CAAC,MAAM,CAAC,aAAa,CAAC,CAAC,CAAC;IACpG,KAAK,CAAC,UAAU,CAAC,OAAO;QACvB,MAAM,cAAc,CAAc,qBAAqB,CAAC,EAAE,EAAE,OAAO,CAAC,CAAC;IACtE,CAAC;CACD,CAAC,CAAC;AAEH,eAAe,gBAAgB,CAAC"}

package/dist/voice/client.d.ts

@@ -0,0 +1,38 @@
+import { BaseRuntimeClient } from '@vibesdotdev/kernel/client';
+import type { AgentVoiceDescriptor } from './agent-voice.descriptor.ts';
+import type { VoicePipelineDeps } from './pipeline.ts';
+import { VoiceSession } from './session.ts';
+export declare const voiceClientDescriptor: {
+    id: string;
+    kind: "runtime/client";
+    name?: string | undefined;
+    tags?: string[] | undefined;
+    hardware?: string[] | undefined;
+    enabled?: boolean | undefined;
+    priority?: number | undefined;
+    capabilityId?: string | undefined;
+    requiredContext?: string[] | undefined;
+    optionalContext?: string[] | undefined;
+    config?: unknown;
+    retention?: string | {
+        reason: string;
+        maxAgeDays?: number | undefined;
+        maxEntries?: number | undefined;
+        maxBytes?: number | undefined;
+        cascadeFrom?: string | undefined;
+        ageField?: string | undefined;
+        enforcedBy?: string | undefined;
+    } | undefined;
+    configManifestId?: string | undefined;
+    baseUrl?: string | undefined;
+    description?: string | undefined;
+};
+/** Owns voice sessions for one runtime and closes them with that runtime. */
+export declare class VoiceClient extends BaseRuntimeClient {
+    private readonly sessions;
+    private closed;
+    createSession(descriptor: AgentVoiceDescriptor, deps: VoicePipelineDeps, systemPrompt?: string): VoiceSession;
+    protected onDispose(): void;
+}
+export default VoiceClient;
+//# sourceMappingURL=client.d.ts.map

package/dist/voice/client.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"client.d.ts","sourceRoot":"","sources":["../../src/voice/client.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,iBAAiB,EAAiC,MAAM,4BAA4B,CAAC;AAE9F,OAAO,KAAK,EAAE,oBAAoB,EAAE,MAAM,6BAA6B,CAAC;AACxE,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,eAAe,CAAC;AACvD,OAAO,EAAE,YAAY,EAAE,MAAM,cAAc,CAAC;AAE5C,eAAO,MAAM,qBAAqB;;;;;;;;;;;;;;;;;;;;;;;;CAEhC,CAAC;AAEH,6EAA6E;AAC7E,qBAAa,WAAY,SAAQ,iBAAiB;IACjD,OAAO,CAAC,QAAQ,CAAC,QAAQ,CAA2B;IACpD,OAAO,CAAC,MAAM,CAAS;IAEvB,aAAa,CAAC,UAAU,EAAE,oBAAoB,EAAE,IAAI,EAAE,iBAAiB,EAAE,YAAY,CAAC,EAAE,MAAM,GAAG,YAAY;cAW1F,SAAS,IAAI,IAAI;CAKpC;AAED,eAAe,WAAW,CAAC"}

package/dist/voice/client.js

@@ -0,0 +1,29 @@
+import { BaseRuntimeClient, RuntimeClientDescriptorSchema } from '@vibesdotdev/kernel/client';
+import { CapabilityError } from '@vibesdotdev/kernel';
+import { VoiceSession } from "./session.js";
+export const voiceClientDescriptor = RuntimeClientDescriptorSchema.parse({
+    id: 'agent-voice', kind: 'runtime/client', name: 'Agent Voice'
+});
+/** Owns voice sessions for one runtime and closes them with that runtime. */
+export class VoiceClient extends BaseRuntimeClient {
+    sessions = new Set();
+    closed = false;
+    createSession(descriptor, deps, systemPrompt) {
+        if (this.closed || this.runtime.lifecycle.signal.aborted) {
+            throw new CapabilityError('The voice runtime is closed.', 'UNAVAILABLE', {
+                reason: 'voice-session-closed', retryable: false
+            });
+        }
+        const session = new VoiceSession(this.runtime, descriptor, deps, systemPrompt, () => this.sessions.delete(session));
+        this.sessions.add(session);
+        return session;
+    }
+    onDispose() {
+        this.closed = true;
+        for (const session of this.sessions)
+            session.close();
+        this.sessions.clear();
+    }
+}
+export default VoiceClient;
+//# sourceMappingURL=client.js.map

package/dist/voice/client.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"client.js","sourceRoot":"","sources":["../../src/voice/client.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,iBAAiB,EAAE,6BAA6B,EAAE,MAAM,4BAA4B,CAAC;AAC9F,OAAO,EAAE,eAAe,EAAE,MAAM,qBAAqB,CAAC;AAGtD,OAAO,EAAE,YAAY,EAAE,MAAM,cAAc,CAAC;AAE5C,MAAM,CAAC,MAAM,qBAAqB,GAAG,6BAA6B,CAAC,KAAK,CAAC;IACxE,EAAE,EAAE,aAAa,EAAE,IAAI,EAAE,gBAAgB,EAAE,IAAI,EAAE,aAAa;CAC9D,CAAC,CAAC;AAEH,6EAA6E;AAC7E,MAAM,OAAO,WAAY,SAAQ,iBAAiB;IAChC,QAAQ,GAAG,IAAI,GAAG,EAAgB,CAAC;IAC5C,MAAM,GAAG,KAAK,CAAC;IAEvB,aAAa,CAAC,UAAgC,EAAE,IAAuB,EAAE,YAAqB;QAC7F,IAAI,IAAI,CAAC,MAAM,IAAI,IAAI,CAAC,OAAO,CAAC,SAAS,CAAC,MAAM,CAAC,OAAO,EAAE,CAAC;YAC1D,MAAM,IAAI,eAAe,CAAC,8BAA8B,EAAE,aAAa,EAAE;gBACxE,MAAM,EAAE,sBAAsB,EAAE,SAAS,EAAE,KAAK;aAChD,CAAC,CAAC;QACJ,CAAC;QACD,MAAM,OAAO,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,OAAO,EAAE,UAAU,EAAE,IAAI,EAAE,YAAY,EAAE,GAAG,EAAE,CAAC,IAAI,CAAC,QAAQ,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC;QACpH,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC;QAC3B,OAAO,OAAO,CAAC;IAChB,CAAC;IAEkB,SAAS;QAC3B,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC;QACnB,KAAK,MAAM,OAAO,IAAI,IAAI,CAAC,QAAQ;YAAE,OAAO,CAAC,KAAK,EAAE,CAAC;QACrD,IAAI,CAAC,QAAQ,CAAC,KAAK,EAAE,CAAC;IACvB,CAAC;CACD;AAED,eAAe,WAAW,CAAC"}

package/dist/voice/index.d.ts

@@ -0,0 +1,14 @@
+/**
+ * @vibesdotdev/agent-execution/voice
+ *
+ * Half-duplex voice pipeline for the `agent/voice` runtime kind.
+ * STT -> text -> NativeToolsAgentLoop (text-only) -> text -> TTS.
+ */
+export { VOICE_KIND_ID, VOICE_AUDIO_FORMATS, VoiceAudioFormatSchema, AgentVoiceDescriptorSchema, createAgentVoiceDescriptor, type VoiceAudioFormat, type AgentVoiceDescriptor } from './agent-voice.descriptor.ts';
+export { VoiceClient, voiceClientDescriptor } from './client.ts';
+export { VoiceSession } from './session.ts';
+export { agentVoiceKind } from './agent-voice.kind.ts';
+export { consumerAgentVoiceDescriptor } from './agent-voice.consumer.descriptor.ts';
+export { agentVoicePlugin, default } from './agent-voice.plugin.ts';
+export { runVoiceTurn, pcm16Rms, SilenceSegmenter, VoiceMetadataSchema, VoiceTurnAudioSchema, VoiceTranscriptSchema, VoiceReplySchema, VoiceSynthesisSchema, VoiceHistoryMessageSchema, VoiceSystemPromptSchema, VoiceTurnResultSchema, VoiceTurnTimingSchema, parseVoiceSystemPrompt, type VoicePipelineDeps, type VoiceMetadata, type VoiceTurnAudio, type VoiceTranscript, type VoiceReply, type VoiceSynthesis, type VoiceHistoryMessage, type VoiceTurnResult } from './pipeline.ts';
+//# sourceMappingURL=index.d.ts.map

package/dist/voice/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/voice/index.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAEH,OAAO,EACN,aAAa,EACb,mBAAmB,EACnB,sBAAsB,EACtB,0BAA0B,EAC1B,0BAA0B,EAC1B,KAAK,gBAAgB,EACrB,KAAK,oBAAoB,EACzB,MAAM,6BAA6B,CAAC;AAErC,OAAO,EAAE,WAAW,EAAE,qBAAqB,EAAE,MAAM,aAAa,CAAC;AACjE,OAAO,EAAE,YAAY,EAAE,MAAM,cAAc,CAAC;AAE5C,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AACvD,OAAO,EACN,4BAA4B,EAC5B,MAAM,sCAAsC,CAAC;AAC9C,OAAO,EAAE,gBAAgB,EAAE,OAAO,EAAE,MAAM,yBAAyB,CAAC;AAEpE,OAAO,EACN,YAAY,EACZ,QAAQ,EACR,gBAAgB,EAChB,mBAAmB,EACnB,oBAAoB,EACpB,qBAAqB,EACrB,gBAAgB,EAChB,oBAAoB,EACpB,yBAAyB,EACzB,uBAAuB,EACvB,qBAAqB,EACrB,qBAAqB,EACrB,sBAAsB,EACtB,KAAK,iBAAiB,EACtB,KAAK,aAAa,EAClB,KAAK,cAAc,EACnB,KAAK,eAAe,EACpB,KAAK,UAAU,EACf,KAAK,cAAc,EACnB,KAAK,mBAAmB,EACxB,KAAK,eAAe,EACpB,MAAM,eAAe,CAAC"}

package/dist/voice/index.js

@@ -0,0 +1,14 @@
+/**
+ * @vibesdotdev/agent-execution/voice
+ *
+ * Half-duplex voice pipeline for the `agent/voice` runtime kind.
+ * STT -> text -> NativeToolsAgentLoop (text-only) -> text -> TTS.
+ */
+export { VOICE_KIND_ID, VOICE_AUDIO_FORMATS, VoiceAudioFormatSchema, AgentVoiceDescriptorSchema, createAgentVoiceDescriptor } from "./agent-voice.descriptor.js";
+export { VoiceClient, voiceClientDescriptor } from "./client.js";
+export { VoiceSession } from "./session.js";
+export { agentVoiceKind } from "./agent-voice.kind.js";
+export { consumerAgentVoiceDescriptor } from "./agent-voice.consumer.descriptor.js";
+export { agentVoicePlugin, default } from "./agent-voice.plugin.js";
+export { runVoiceTurn, pcm16Rms, SilenceSegmenter, VoiceMetadataSchema, VoiceTurnAudioSchema, VoiceTranscriptSchema, VoiceReplySchema, VoiceSynthesisSchema, VoiceHistoryMessageSchema, VoiceSystemPromptSchema, VoiceTurnResultSchema, VoiceTurnTimingSchema, parseVoiceSystemPrompt } from "./pipeline.js";
+//# sourceMappingURL=index.js.map

package/dist/voice/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/voice/index.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAEH,OAAO,EACN,aAAa,EACb,mBAAmB,EACnB,sBAAsB,EACtB,0BAA0B,EAC1B,0BAA0B,EAG1B,MAAM,6BAA6B,CAAC;AAErC,OAAO,EAAE,WAAW,EAAE,qBAAqB,EAAE,MAAM,aAAa,CAAC;AACjE,OAAO,EAAE,YAAY,EAAE,MAAM,cAAc,CAAC;AAE5C,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AACvD,OAAO,EACN,4BAA4B,EAC5B,MAAM,sCAAsC,CAAC;AAC9C,OAAO,EAAE,gBAAgB,EAAE,OAAO,EAAE,MAAM,yBAAyB,CAAC;AAEpE,OAAO,EACN,YAAY,EACZ,QAAQ,EACR,gBAAgB,EAChB,mBAAmB,EACnB,oBAAoB,EACpB,qBAAqB,EACrB,gBAAgB,EAChB,oBAAoB,EACpB,yBAAyB,EACzB,uBAAuB,EACvB,qBAAqB,EACrB,qBAAqB,EACrB,sBAAsB,EAStB,MAAM,eAAe,CAAC"}

package/dist/voice/pipeline.d.ts

@@ -0,0 +1,66 @@
+/**
+ * Half-duplex voice pipeline — the engine behind the `agent/voice` kind.
+ *
+ * One *turn* = one user utterance → one assistant reply:
+ *
+ *   audio (PCM/encoded) ──STT──▶ transcript ──agent loop──▶ reply text ──TTS──▶ audio
+ *
+ * The pipeline is deliberately **dependency-injected**: STT, the text turn
+ * (agent loop), and TTS are passed in as async functions. The host (the
+ * chat-web voice route) resolves those from the runtime's `ai/generation`
+ * providers + the agent loop, so this package stays workerd-safe — it never
+ * imports `@vibesdotdev/filesystem`, Playwright, or any Node-only generation
+ * executor. It only depends on the runtime + ai *schemas*.
+ *
+ * Server-side silence detection (`SilenceSegmenter`) lives here so both the
+ * WS/POST transport and any future full-duplex transport can reuse the same
+ * turn-boundary logic.
+ */
+import { type RuntimeRef } from '@vibesdotdev/kernel';
+import { type AgentVoiceDescriptor } from './agent-voice.descriptor.ts';
+import { type VoiceTurnAudio, type VoiceHistoryMessage, type VoiceTurnResult, type VoicePipelineDeps } from '../schemas/voice.ts';
+export { VoiceMetadataSchema, VoiceTurnAudioSchema, VoiceTranscriptSchema, VoiceReplySchema, VoiceSynthesisSchema, VoiceHistoryMessageSchema, VoiceHistorySchema, VoiceAbortSignalSchema, VoiceSystemPromptSchema, VoiceTurnTimingSchema, VoiceTurnResultSchema, VoicePipelineDepsSchema, type VoiceMetadata, type VoiceTurnAudio, type VoiceTranscript, type VoiceReply, type VoiceSynthesis, type VoiceHistoryMessage, type VoiceTurnResult, type VoicePipelineDeps } from '../schemas/voice.ts';
+export declare function parseVoiceSystemPrompt<TInput>(value: TInput): string;
+/**
+ * Run a single half-duplex turn under the descriptor's `turnTimeoutMs`, a hard
+ * ceiling over the whole STT→agent→TTS chain. The ceiling is the owning
+ * runtime's bounded wait (RC07): `runtime` is the runtime the host was handed,
+ * so disposing that runtime aborts the turn and
+ * no timer outlives it. Every stage receives the one combined signal (the
+ * caller's, the runtime's and the deadline). Exceeding the ceiling rejects
+ * `CapabilityError` `voice-turn-timeout`.
+ */
+export declare function runVoiceTurn(runtime: RuntimeRef, descriptor: AgentVoiceDescriptor, deps: VoicePipelineDeps, audio: VoiceTurnAudio, history: VoiceHistoryMessage[], externalSignal?: AbortSignal): Promise<VoiceTurnResult>;
+/**
+ * RMS amplitude of a 16-bit little-endian PCM frame, normalized to 0..1.
+ * Used by the silence segmenter. For non-PCM (encoded) frames the caller
+ * should not rely on this — silence detection is a PCM-stream concern.
+ */
+export declare function pcm16Rms(bytes: Uint8Array): number;
+/**
+ * Server-side silence detection that segments an inbound PCM stream into turns.
+ *
+ * Feed frames as they arrive via {@link push}; when trailing silence exceeds
+ * `silenceTimeoutMs` after at least one voiced frame, {@link push} returns the
+ * concatenated utterance bytes (and resets). This is the half-duplex turn
+ * boundary: one completed utterance → one agent turn.
+ */
+export declare class SilenceSegmenter {
+    private readonly frames;
+    private bytesBuffered;
+    private hasVoiced;
+    private silenceAccumulatedMs;
+    private readonly threshold;
+    private readonly silenceTimeoutMs;
+    private readonly bytesPerMs;
+    constructor(descriptor: Pick<AgentVoiceDescriptor, 'silenceThreshold' | 'silenceTimeoutMs' | 'sampleRate'>);
+    /**
+     * Push a PCM frame. Returns the completed utterance bytes when trailing
+     * silence terminates a turn, otherwise `null`.
+     */
+    push(frame: Uint8Array): Uint8Array | null;
+    /** Force-close the current utterance (e.g. transport closed). */
+    flush(): Uint8Array | null;
+    private reset;
+}
+//# sourceMappingURL=pipeline.d.ts.map

package/dist/voice/pipeline.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"pipeline.d.ts","sourceRoot":"","sources":["../../src/voice/pipeline.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;GAiBG;AAGH,OAAO,EAMN,KAAK,UAAU,EAEf,MAAM,qBAAqB,CAAC;AAC7B,OAAO,EAEN,KAAK,oBAAoB,EACzB,MAAM,6BAA6B,CAAC;AAErC,OAAO,EASN,KAAK,cAAc,EAInB,KAAK,mBAAmB,EACxB,KAAK,eAAe,EACpB,KAAK,iBAAiB,EACtB,MAAM,qBAAqB,CAAC;AAC7B,OAAO,EACN,mBAAmB,EACnB,oBAAoB,EACpB,qBAAqB,EACrB,gBAAgB,EAChB,oBAAoB,EACpB,yBAAyB,EACzB,kBAAkB,EAClB,sBAAsB,EACtB,uBAAuB,EACvB,qBAAqB,EACrB,qBAAqB,EACrB,uBAAuB,EACvB,KAAK,aAAa,EAClB,KAAK,cAAc,EACnB,KAAK,eAAe,EACpB,KAAK,UAAU,EACf,KAAK,cAAc,EACnB,KAAK,mBAAmB,EACxB,KAAK,eAAe,EACpB,KAAK,iBAAiB,EACtB,MAAM,qBAAqB,CAAC;AAE7B,wBAAgB,sBAAsB,CAAC,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,CAWpE;AAsCD;;;;;;;;GAQG;AACH,wBAAsB,YAAY,CACjC,OAAO,EAAE,UAAU,EACnB,UAAU,EAAE,oBAAoB,EAChC,IAAI,EAAE,iBAAiB,EACvB,KAAK,EAAE,cAAc,EACrB,OAAO,EAAE,mBAAmB,EAAE,EAC9B,cAAc,CAAC,EAAE,WAAW,GAC1B,OAAO,CAAC,eAAe,CAAC,CA0I1B;AAED;;;;GAIG;AACH,wBAAgB,QAAQ,CAAC,KAAK,EAAE,UAAU,GAAG,MAAM,CAElD;AAcD;;;;;;;GAOG;AACH,qBAAa,gBAAgB;IAC5B,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAoB;IAC3C,OAAO,CAAC,aAAa,CAAK;IAC1B,OAAO,CAAC,SAAS,CAAS;IAC1B,OAAO,CAAC,oBAAoB,CAAK;IACjC,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAS;IACnC,OAAO,CAAC,QAAQ,CAAC,gBAAgB,CAAS;IAC1C,OAAO,CAAC,QAAQ,CAAC,UAAU,CAAS;gBAExB,UAAU,EAAE,IAAI,CAAC,oBAAoB,EAAE,kBAAkB,GAAG,kBAAkB,GAAG,YAAY,CAAC;IAwB1G;;;OAGG;IACH,IAAI,CAAC,KAAK,EAAE,UAAU,GAAG,UAAU,GAAG,IAAI;IA0B1C,iEAAiE;IACjE,KAAK,IAAI,UAAU,GAAG,IAAI;IAe1B,OAAO,CAAC,KAAK;CAMb"}