@autoark-ai/eva-client-sdk-ts 0.0.2-dev

package/dist/index.d.ts

@@ -0,0 +1,406 @@
+import type { MediaTransportsConfig } from "./spi.js";
+// Generated by dts-bundle-generator v9.5.1
+/** JSON-compatible value accepted at public metadata boundaries. */
+export type JsonValue = null | boolean | number | string | readonly JsonValue[] | JsonObject;
+/** Readonly string-keyed JSON object. */
+export interface JsonObject {
+    readonly [key: string]: JsonValue;
+}
+/** A final user or assistant message committed by one agent session. */
+export interface ConversationMessage {
+    /** Stable identifier unique within the agent instance. */
+    readonly id: string;
+    /** Turn identity shared by the corresponding user/assistant messages. */
+    readonly turnId: string;
+    /** Public alpha message role. */
+    readonly role: "user" | "assistant";
+    /** Final text content. */
+    readonly content: string;
+    /** Commit time as Unix epoch milliseconds. */
+    readonly createdAt: number;
+    /** Recursively isolated caller metadata snapshot. */
+    readonly metadata: JsonObject;
+}
+type ErrorSource = "sdk" | "provider" | "gateway" | "media";
+type MediaRole = "audio-input" | "audio-output" | "aec" | "camera";
+type MediaOperation = "start" | "capture" | "stop";
+type MediaErrorReason = "not_configured" | "permission_denied" | "device_unavailable" | "unsupported" | "timeout" | "invalid_data" | "operation_failed";
+/** Sanitized error shape exposed by rejected API calls and `error` events. */
+export interface StructuredError {
+    /** Stable, non-sensitive summary suitable for application logs and UI. */
+    message: string;
+    /** Whether the affected operation/session cannot safely continue without caller action. */
+    fatal: boolean;
+    /** Subsystem that owns this failure. */
+    source: ErrorSource;
+    /** Provider identifier when `source` is `provider` or `gateway`. */
+    provider?: string;
+    /** HTTP status code when a Gateway response supplied one. */
+    statusCode?: number;
+    /** Media role that failed; present only when `source` is `media`. */
+    role?: MediaRole;
+    /** Media lifecycle operation that failed; present only when `source` is `media`. */
+    operation?: MediaOperation;
+    /** Stable media failure reason; present only when `source` is `media`. */
+    reason?: MediaErrorReason;
+}
+interface EvaSdkErrorOptions {
+    fatal?: boolean;
+    cause?: unknown;
+}
+export declare class EvaSdkError extends Error implements StructuredError {
+    readonly fatal: boolean;
+    readonly source: ErrorSource;
+    constructor(message: string, options?: EvaSdkErrorOptions);
+}
+/** Controls whether and how the agent starts a greeting turn. */
+export type GreetingConfig = {
+    /** Disables the automatic greeting turn. */
+    mode: "disabled";
+} | {
+    /** Plays `text` directly as the assistant greeting without calling the LLM. */
+    mode: "static";
+    /** Non-empty assistant text synthesized for the greeting turn. */
+    text: string;
+} | {
+    /** Calls the LLM to generate the assistant greeting. */
+    mode: "dynamic";
+    /** Optional LLM instruction for the greeting; blank or omitted uses the SDK default. */
+    prompt?: string;
+};
+/** Bounded conversation history used as context for later LLM turns. */
+export interface HistoryConfig {
+    /**
+     * Maximum completed user/assistant turn pairs retained as LLM context.
+     * @defaultValue 10
+     * @remarks Must be a positive integer. History is enabled only when `history` is provided.
+     */
+    maxTurns?: number;
+}
+/** Camera capture timing used by the managed dialogue runtime. */
+export interface CameraConfig {
+    /**
+     * Maximum duration of one still-image capture before the turn falls back to text-only.
+     * @defaultValue 1500
+     * @remarks Must be a finite positive integer in milliseconds. This is separate from the
+     * fixed internal cancellation-settlement deadline.
+     */
+    captureTimeoutMs?: number;
+}
+interface RuntimeConfig {
+    /**
+     * System instruction prepended to each LLM request.
+     * @defaultValue An empty string.
+     */
+    systemPrompt?: string;
+    /**
+     * Optional greeting behavior executed when the agent starts.
+     * @defaultValue `{ mode: "disabled" }`
+     */
+    greeting?: GreetingConfig;
+    /**
+     * Enables bounded multi-turn LLM context when provided.
+     * @remarks Omit this field to keep turns independent.
+     */
+    history?: HistoryConfig;
+    /**
+     * Optional camera capture behavior.
+     * @remarks Camera remains disabled by default even when this field and `transports.camera`
+     * are present; call `setCameraCaptureEnabled(true)` to acquire the camera session.
+     */
+    camera?: CameraConfig;
+    /**
+     * Complete audio input, output, AEC, and optional camera role set owned by the runtime after construction.
+     * @remarks Omit this field for a text-only agent. Do not drive these roles concurrently with the agent.
+     */
+    transports?: MediaTransportsConfig;
+    /**
+     * JSON-compatible metadata attached to agent-produced events and conversation messages.
+     * @defaultValue An empty object.
+     * @remarks The SDK clones this value at the public boundary.
+     */
+    metadata?: JsonObject;
+}
+/** Public managed configuration for one Eva voice-dialogue agent instance. */
+export interface EvaVoiceDialogueAgentConfig extends RuntimeConfig {
+    /**
+     * Gateway credential shared by the built-in ASR, LLM, and TTS stages.
+     * @remarks The SDK does not expose this value through events, metadata, or configuration getters.
+     */
+    apiKey: string;
+    /** Managed automatic speech recognition configuration. */
+    asr: {
+        /** Gateway ASR model identifier forwarded as the wire `model` field. */
+        model: string;
+        /**
+         * Target PCM sample rate sent to the configured ASR model, in Hz.
+         * @remarks Must be a finite positive integer. The SDK resamples at the ASR boundary and does not infer model compatibility.
+         */
+        sampleRate: number;
+    };
+    /** Managed text-to-speech configuration. */
+    tts: {
+        /** Gateway TTS model identifier forwarded as the wire `model` field. */
+        model: string;
+        /** Optional model-specific voice identifier. */
+        voice?: string;
+        /** Optional speaking speed forwarded as the Gateway `speed` value. */
+        speakingRate?: number;
+        /** Optional pitch multiplier forwarded as the Gateway `pitch_rate` value. */
+        pitch?: number;
+        /**
+         * Requested TTS PCM sample rate in Hz.
+         * @defaultValue 16000
+         * @remarks Must be a finite positive integer when provided. This controls synthesis output, not the playback device sample rate.
+         */
+        sampleRate?: number;
+    };
+    /** Managed large-language-model configuration. */
+    llm: {
+        /** Gateway LLM model identifier forwarded as the wire `model` field. */
+        model: string;
+        /** Optional sampling temperature forwarded to the selected LLM. */
+        temperature?: number;
+        /** Optional maximum generated token count forwarded as `max_tokens`. */
+        maxTokens?: number;
+    };
+    /**
+     * Optional local Silero voice-activity-detection tuning.
+     * @remarks Required when `transports.input` is configured.
+     */
+    vad?: {
+        /**
+         * Speech probability threshold used to enter the speaking state.
+         * @defaultValue 0.5
+         * @remarks Must be greater than 0 and less than or equal to 1.
+         */
+        sensitivity?: number;
+        /**
+         * Continuous silence required before speech is considered stopped, in milliseconds.
+         * @defaultValue 200
+         * @remarks Must be finite and greater than 0. The Silero stage rounds up to 32 ms frames.
+         */
+        silenceThresholdMs?: number;
+    };
+}
+/** The closed alpha event catalog. This is type-only; no runtime enum is exported. */
+export type AgentEventType = "speech.started" | "image.captured" | "speech.stopped" | "transcript.partial" | "transcript.final" | "interruption" | "reply.started" | "reply.partial" | "reply.final" | "playback.started" | "playback.stopped" | "turn.latency" | "error";
+/** Shared envelope fields projected from the language-neutral frame model. */
+export interface AgentEventEnvelope {
+    /** Workflow identity such as speech, manual-text, or greeting. */
+    readonly streamId: string;
+    /** Turn identity; optional only for an error whose turn cannot be determined. */
+    readonly turnId?: string;
+    /** Monotonic sequence within one stream. */
+    readonly sequence?: number;
+    /** Whether this event frame is an incremental fragment. */
+    readonly partial: boolean;
+    /** Whether this event frame is complete. */
+    readonly final: boolean;
+    /** Optional event time in Unix epoch milliseconds. */
+    readonly timestamp?: number;
+    /** Sanitized frame metadata; secrets and raw provider payloads are removed. */
+    readonly metadata: Readonly<Record<string, unknown>>;
+    /** Optional internal-frame correlation identity. */
+    readonly frameId?: string;
+}
+interface TurnEventEnvelope extends Omit<AgentEventEnvelope, "turnId" | "partial" | "final"> {
+    readonly turnId: string;
+}
+export interface SpeechStartedEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "speech.started";
+    /** Speech start is a discrete event, not a fragment. */
+    readonly partial: false;
+    /** Speech start is complete when emitted. */
+    readonly final: true;
+}
+/** Sanitized local camera snapshot summary. This does not mean the LLM accepted or used the image. */
+export interface ImageCapturedEvent extends TurnEventEnvelope {
+    /** Public event discriminator for a locally completed and validated snapshot. */
+    readonly type: "image.captured";
+    /** Image capture is a discrete event, not an incremental fragment. */
+    readonly partial: false;
+    /** The local capture result is complete when this event is emitted. */
+    readonly final: true;
+    /**
+     * Non-sensitive snapshot metadata safe for logs and UI.
+     * @remarks This object intentionally excludes image bytes, base64 data, device identifiers,
+     * and any claim that the Gateway or model accepted or used the image.
+     */
+    readonly image: {
+        /** Actual MIME type reported by the validated snapshot, for example `image/png`. */
+        readonly mimeType: string;
+        /** Captured image width in pixels. */
+        readonly width: number;
+        /** Captured image height in pixels. */
+        readonly height: number;
+        /** Encoded image payload size in bytes; the payload itself is not exposed by the event. */
+        readonly sizeBytes: number;
+        /** Local capture duration in milliseconds, measured by the SDK runtime. */
+        readonly captureMs: number;
+    };
+}
+export interface SpeechStoppedEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "speech.stopped";
+    /** Speech stop is a discrete event, not a fragment. */
+    readonly partial: false;
+    /** Speech stop is complete when emitted. */
+    readonly final: true;
+}
+export interface TranscriptPartialEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "transcript.partial";
+    /** A partial transcript is an incremental fragment. */
+    readonly partial: true;
+    /** A partial transcript is not the final transcript. */
+    readonly final: false;
+    /** Current incremental speech transcript text. */
+    readonly text: string;
+    /** Partial transcripts are emitted only for captured speech. */
+    readonly source: "speech";
+}
+export interface TranscriptFinalEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "transcript.final";
+    /** A final transcript is not an incremental fragment. */
+    readonly partial: false;
+    /** Marks the complete user transcript for the turn. */
+    readonly final: true;
+    /** Complete final user transcript text. */
+    readonly text: string;
+    /** Whether the turn originated from speech or submitText(). */
+    readonly source: "speech" | "text";
+}
+export interface InterruptionEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "interruption";
+    /** Interruption is a discrete event, not a fragment. */
+    readonly partial: false;
+    /** Interruption is complete when emitted. */
+    readonly final: true;
+    /** User action that superseded the active turn. */
+    readonly reason: "user_speech" | "manual_text";
+}
+export interface ReplyStartedEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "reply.started";
+    /** Reply start is a discrete event, not a fragment. */
+    readonly partial: false;
+    /** Reply start is complete when emitted. */
+    readonly final: true;
+}
+export interface ReplyPartialEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "reply.partial";
+    /** A reply delta is an incremental fragment. */
+    readonly partial: true;
+    /** A reply delta is not the final reply. */
+    readonly final: false;
+    /** Newly appended reply delta. */
+    readonly text: string;
+}
+export interface ReplyFinalEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "reply.final";
+    /** A final reply is not an incremental fragment. */
+    readonly partial: false;
+    /** Marks the complete assistant reply for the turn. */
+    readonly final: true;
+    /** Complete final reply for the turn. */
+    readonly text: string;
+}
+export interface PlaybackStartedEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "playback.started";
+    /** Playback start is a discrete event, not a fragment. */
+    readonly partial: false;
+    /** Playback start is complete when emitted. */
+    readonly final: true;
+}
+export interface PlaybackStoppedEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "playback.stopped";
+    /** Playback stop is a discrete event, not a fragment. */
+    readonly partial: false;
+    /** Playback stop is complete when emitted. */
+    readonly final: true;
+}
+/** Optional duration measurements for one completed turn. */
+export interface TurnLatencyStages {
+    /** Voice activity detection duration in milliseconds. */
+    readonly vadMs?: number;
+    /** Speech recognition duration in milliseconds. */
+    readonly asrMs?: number;
+    /** Time to the first LLM token in milliseconds. */
+    readonly llmFirstTokenMs?: number;
+    /** Time to the first synthesized audio chunk in milliseconds. */
+    readonly ttsFirstAudioMs?: number;
+    /** Audio playback duration in milliseconds. */
+    readonly playbackMs?: number;
+}
+/** Latency summary correlated to one turn. */
+export interface TurnLatencyPayload {
+    /** Turn measured by this payload. */
+    readonly turnId: string;
+    /** Optional end-to-end turn duration in milliseconds. */
+    readonly totalMs?: number;
+    /** Available per-stage measurements. */
+    readonly stages: TurnLatencyStages;
+}
+export interface TurnLatencyEvent extends TurnEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "turn.latency";
+    /** Latency is a discrete event, not a fragment. */
+    readonly partial: false;
+    /** Latency is complete when emitted. */
+    readonly final: true;
+    /** Turn latency measurements. */
+    readonly latency: TurnLatencyPayload;
+}
+export interface AgentErrorEvent extends AgentEventEnvelope {
+    /** Public event discriminator. */
+    readonly type: "error";
+    /** Error is a discrete event, not a fragment. */
+    readonly partial: false;
+    /** Error is complete when emitted. */
+    readonly final: true;
+    /** Sanitized SDK/provider failure safe for application handling. */
+    readonly error: StructuredError;
+}
+/** Discriminated union for every event emitted by the alpha Facade. */
+export type AgentEvent = SpeechStartedEvent | ImageCapturedEvent | SpeechStoppedEvent | TranscriptPartialEvent | TranscriptFinalEvent | InterruptionEvent | ReplyStartedEvent | ReplyPartialEvent | ReplyFinalEvent | PlaybackStartedEvent | PlaybackStoppedEvent | TurnLatencyEvent | AgentErrorEvent;
+/** Synchronous observer invoked for each projected public event. */
+export type AgentEventListener = (event: AgentEvent) => void;
+/** Idempotent function that removes one event listener. */
+export type Unsubscribe = () => void;
+/** Optional identity and caller metadata for one manual-text turn. */
+export interface SubmitTextOptions {
+    readonly turnId?: string;
+    readonly metadata?: JsonObject;
+}
+/** Minimal public control and observation surface for one dialogue session. */
+export interface EvaVoiceDialogueAgent {
+    start(): Promise<void>;
+    submitText(text: string, options?: SubmitTextOptions): Promise<void>;
+    setAudioInputEnabled(enabled: boolean): Promise<void>;
+    /**
+     * Enables or disables the optional continuously held camera capture session.
+     *
+     * @param enabled - `true` acquires the configured camera and keeps its session ready for
+     * speech-triggered snapshots; `false` prevents new captures, cancels an in-flight capture,
+     * and waits for camera resources to be released.
+     * @returns A Promise that resolves after the requested camera state is fully ready or released.
+     * Repeated calls with the same value share the same in-flight result.
+     * @remarks Camera capture is disabled by default. Enabling rejects when no camera source is
+     * configured or acquisition fails. This control does not enable/disable audio input or TTS.
+     */
+    setCameraCaptureEnabled(enabled: boolean): Promise<void>;
+    setTtsEnabled(enabled: boolean): Promise<void>;
+    getMessages(): readonly ConversationMessage[];
+    onEvent(listener: AgentEventListener): Unsubscribe;
+    stop(): Promise<void>;
+}
+export declare function createEvaVoiceDialogueAgent(config: EvaVoiceDialogueAgentConfig): EvaVoiceDialogueAgent;
+export {};

package/dist/index.js

@@ -0,0 +1,3 @@
+function C(e){if(!Pe(e))throw new TypeError("Metadata must be a JSON-compatible object");return xe(e,new Set)}function Re(e,t){if(e===null||typeof e=="boolean"||typeof e=="string")return e;if(typeof e=="number"){if(!Number.isFinite(e))throw new TypeError("Metadata numbers must be finite");return e}if(Array.isArray(e))return Ie(e,t,()=>e.map(r=>Re(r,t)));if(Pe(e))return xe(e,t);throw new TypeError("Metadata must contain only JSON-compatible values")}function xe(e,t){return Ie(e,t,()=>{if(Reflect.ownKeys(e).some(n=>typeof n!="string"))throw new TypeError("Metadata object keys must be strings");let r={};for(let[n,a]of Object.entries(e))r[n]=Re(a,t);return r})}function Ie(e,t,r){if(t.has(e))throw new TypeError("Metadata must not contain cycles");t.add(e);try{return r()}finally{t.delete(e)}}function Pe(e){if(typeof e!="object"||e===null||Array.isArray(e))return!1;let t=Object.getPrototypeOf(e);return t===Object.prototype||t===null}var h=class extends Error{fatal;source="sdk";constructor(t,r={}){super(t,{cause:r.cause}),this.name="EvaSdkError",this.fatal=r.fatal??!0}},N=class extends h{provider;source="provider";constructor(t,r){super(t,r),this.name="StageProviderError",this.provider=r.provider}},W=class extends h{provider;statusCode;source="gateway";constructor(t,r){super(t,r),this.name="GatewayAccessError",this.provider=r.provider,r.statusCode!==void 0&&(this.statusCode=r.statusCode)}},O=class extends h{role;operation;reason;source="media";constructor(t,r){super(t,r),this.name="MediaIoError",this.role=r.role,this.operation=r.operation,this.reason=r.reason}};function w(e,t={}){return e instanceof h?e:new h(t.message??"SDK operation failed",{fatal:t.fatal??!0,cause:e})}function E(e,t){return e instanceof h?e:new N(t.message??"Stage provider failed",{provider:t.provider,fatal:t.fatal??!0,cause:e})}function L(e,t){if(e instanceof h)return e;let r={provider:t.provider,fatal:t.fatal??!0,cause:e};return t.statusCode!==void 0&&(r.statusCode=t.statusCode),new W(t.message??vt(t.statusCode),r)}function vt(e){return e===void 0?"Gateway request failed":`Gateway request failed with status ${e}`}var St=new Set(["pcm_s16le"]);function ie(e,t){if(!St.has(e.format))throw new h("Unsupported audio format",{fatal:!0});return{kind:"audio.input",streamId:t.streamId,...t.turnId!==void 0?{turnId:t.turnId}:{},...t.sequence!==void 0?{sequence:t.sequence}:{},partial:!0,final:!1,...t.timestamp!==void 0?{timestamp:t.timestamp}:{},metadata:t.metadata??{},audio:e.data,sampleRate:e.sampleRate,channels:e.channels}}function oe(e){return{data:e.audio,sampleRate:e.sampleRate,channels:e.channels,format:"pcm_s16le"}}function Fe(){let e,t=()=>{let r=e;if(r!==void 0)return e=void 0,r.controller.abort(),r};return{begin(r){t();let n=new AbortController;return e={...r,controller:n,signal:n.signal},e},current(){return e},cancel:t,complete(r){return e!==r?!1:(e=void 0,!0)},isCurrent(r){return e===r&&!r.signal.aborted},stop:t}}function ke(){let e="";return{push(t){if(t.length===0)return[];e+=t;let r=wt(e);return e=r.rest,r.sentences},flush(){let t=e.trim();return e="",t.length===0?[]:[t]},clear(){e=""}}}var bt=new Set(["\u3002","\uFF01","\uFF1F","!","?","\uFF1B",";","\u2026"]),At=new Set(['"',"'","\u201D","\u2019",")","\uFF09","]","\u3011","}","\u300B","\u300D","\u300F"]),Et=/(?:Mr|Mrs|Ms|Dr|Prof|Sr|Jr|St|vs|etc|e\.g|i\.e)\.$/i;function wt(e){let t=[],r=0,n=0;for(;n<e.length;){if(!Tt(e,n)){n+=1;continue}let a=n+1;for(;a<e.length&&At.has(e[a]);)a+=1;let i=a;for(;i<e.length&&/\s/u.test(e[i]);)i+=1;if(i>=e.length)break;let s=e.slice(r,a).trim();s.length>0&&t.push(s),r=i,n=i}return{sentences:t,rest:e.slice(r)}}function Tt(e,t){let r=e[t];if(bt.has(r))return!(r==="\u2026"&&e[t+1]==="\u2026");if(r!==".")return!1;let n=e[t-1],a=e[t+1];return n!==void 0&&a!==void 0&&/\d/u.test(n)&&/\d/u.test(a)||a==="."?!1:!Et.test(e.slice(0,t+1))}function Me(e){let t=e,r=0,n=()=>{let o,l=new Promise(m=>{o=m});return{id:r,queue:[],invalidated:l,invalidate:o,currentController:void 0,pump:void 0}},a=n(),i=!1,s,u=()=>{let o=a;r+=1,o.queue.length=0,o.currentController?.abort(),o.invalidate(),a=n()},c=()=>{i=!0,u()};t.parentSignal.aborted?c():t.parentSignal.addEventListener("abort",c,{once:!0});let d=o=>{o.pump!==void 0||o.queue.length===0||(o.pump=p(o).catch(l=>{o===a&&(s=l,i=!0,u())}).finally(()=>{o.pump=void 0,o===a&&o.queue.length>0&&d(o)}))};async function p(o){for(;o.queue.length>0&&!t.parentSignal.aborted;){let l=o.queue.shift(),m=new AbortController;o.currentController=m;let v={signal:m.signal,isCurrent:()=>!t.parentSignal.aborted&&!m.signal.aborted&&o===a&&o.id===r};try{await t.process(l,v)}finally{o.currentController===m&&(o.currentController=void 0)}}}return{enqueue(o){i||t.parentSignal.aborted||o.trim().length===0||(a.queue.push(o),d(a))},clearAndAbort:u,async close(){i=!0;let o=a,l=o.pump;if(l!==void 0&&await Promise.race([l,o.invalidated]),t.parentSignal.removeEventListener("abort",c),s!==void 0)throw s}}}function Oe(e){let t=e,r=!1,n;return{start(){r||n!==void 0||(t.onStarted(),r=!0)},async close(){if(!r){n!==void 0&&await n;return}r=!1,n=Promise.resolve(t.onStopped()).finally(()=>{n=void 0}),await n},isActive(){return r}}}function Le(e){let t=e,r=t.now??Rt,n=r(),a,i,s=t.source==="text"||t.source==="greeting"?n:void 0,u,c,d,p=!1,o={},l=()=>{let m=t.source==="text"?0:o.vadMs??j(n,a),v=t.source==="text"?0:o.asrMs??j(i??a,s),f=o.llmFirstTokenMs??j(s,u),g=o.ttsFirstAudioMs??j(u,c),b=o.playbackMs??j(c,d),y={};D(y,"vadMs",m),D(y,"asrMs",v),D(y,"llmFirstTokenMs",f),D(y,"ttsFirstAudioMs",g),D(y,"playbackMs",b),Object.freeze(y);let S=[m,v,f,g],A=S.every(P=>P!==void 0)?S.reduce((P,ae)=>P+ae,0):void 0;return Object.freeze({turnId:t.turnId,...A!==void 0?{totalMs:A}:{},stages:y})};return{markVadStarted(){a??=r()},markAsrStarted(){i??=r()},markAsrFinal(){s??=r()},markLlmFirstToken(){u??=r()},markTtsFirstAudio(){c??=r()},markPlaybackStarted(){d??=r()},recordStageMetadata(m,v){let f=Ct[m],g=v[f];typeof g=="number"&&Number.isFinite(g)&&g>=0&&(o[f]=Math.round(g))},snapshot:l,takeSnapshot(){if(p)return;let m=l();if(Object.keys(m.stages).length!==0)return p=!0,m}}}var Ct={vad:"vadMs",asr:"asrMs",llm:"llmFirstTokenMs",tts:"ttsFirstAudioMs"};function Rt(){return typeof performance>"u"?Date.now():performance.now()}function j(e,t){if(!(e===void 0||t===void 0))return Math.max(0,Math.round(t-e))}function D(e,t,r){r!==void 0&&(e[t]=r)}function De(e,t={}){let r=xt(t.preSpeechMs),n=It(t.maxUtteranceMs),a=[],i=[],s=new Set,u=0,c,d=0,p=0,o=!1,l,m=()=>{for(let f of s)f();s.clear()},v=f=>{l??=new h(f,{fatal:!0}),o=!0,c=void 0,i.length=0,m()};return{async*vadAudio(){try{for await(let f of e){if(l!==void 0)throw l;let g=je(f);for(a.push(f),u+=g;a.length>1;){let b=a[0],y=je(b);if(u-y<r)break;a.shift(),u-=y}if(c!==void 0&&(c.frames.push(f),p+=g,p>n))throw v("VAD utterance exceeded max duration"),l;yield f}}catch(f){throw l??=f,o=!0,m(),f}},start(f,g=d+1){d=g,c={turnId:f,generation:g,frames:[...a]},p=u},stop(){c!==void 0&&c.frames.length>0&&(i.length=0,i.push(c)),c=void 0,p=0,a.length=0,u=0,m()},async*utterances(){for(;;){let f=i.pop();if(i.length=0,f!==void 0){yield{turnId:f.turnId,generation:f.generation,audio:Pt(f.frames)};continue}if(l!==void 0)throw l;if(o)return;await new Promise(g=>s.add(g))}},close(){o||(o=!0,c=void 0,m())}}}function xt(e){return Number.isFinite(e)&&e!==void 0&&e>=0?e:200}function It(e){return Number.isFinite(e)&&e!==void 0&&e>0?e:6e4}function je(e){return e.sampleRate<=0||e.channels<=0?0:e.audio.byteLength/2/e.channels/e.sampleRate*1e3}async function*Pt(e){for(let t of e)yield t}var Ft=1500,V=class extends Error{turnId;generation;constructor(t,r,n){super("Camera capture cancellation failed",{cause:t}),this.name="CameraCaptureSettlementError",this.turnId=r,this.generation=n,Object.defineProperty(this,"mediaError",{value:t,enumerable:!1,configurable:!1,writable:!1})}},q=class{source;now;settlementDeadlineMs;onFault;controlTail=Promise.resolve();acceptedControl=Promise.resolve();stopOperation;running=!1;stopping=!1;acceptedEnabled=!1;acceptedRequestId=0;active=!1;sessionIdentity=0;sessionController;pendingStart;pendingCapture;fault;constructor(t){this.source=t.source,this.now=t.now??Date.now,this.settlementDeadlineMs=t.cancellationSettlementDeadlineMs??Ft,this.onFault=t.onFault}isActive(){return this.active&&!this.stopping&&this.fault===void 0}currentFault(){return this.fault}setEnabled(t){if(this.stopping)return Promise.reject(this.faultedControlError(t?"start":"stop"));if(this.fault!==void 0)return Promise.reject(this.faultedControlError(t?"start":"stop"));if(this.acceptedEnabled===t)return this.acceptedControl;this.acceptedEnabled=t;let r=++this.acceptedRequestId;if(t||this.abortPendingWork(),!this.running)return this.acceptedControl=Promise.resolve(),this.acceptedControl;let n=this.enqueueControl(()=>this.applyEnabled(t,r));return this.acceptedControl=n.catch(a=>{throw this.acceptedRequestId===r&&(this.acceptedEnabled=!1),a}),this.acceptedControl}async startRuntime(){if(!this.running&&(this.running=!0,this.stopping=!1,!!this.acceptedEnabled))try{this.acceptedControl=this.enqueueControl(()=>this.applyEnabled(!0,this.acceptedRequestId)),await this.acceptedControl}catch(t){throw this.acceptedEnabled=!1,t}}async stopRuntime(){if(this.stopping)return this.stopOperation??Promise.resolve();this.stopping=!0,this.running=!1,this.acceptedEnabled=!1,this.abortPendingWork();let t=this.enqueueControl(async()=>{let r=this.source;if(r!==void 0)try{await this.stopSourceAfterCaptureSettlement(r,!0)}catch(n){throw this.markFault("stop","operation_failed",n)}finally{this.active=!1,this.sessionController=void 0}});return this.stopOperation=t,t}async beginCapture(t,r,n){if(await this.cancelPendingCaptureAndWait(),!this.isActive()||this.source===void 0)return;let a=new AbortController,i=Mt(),s=this.now(),u={turnId:t,generation:r},d=Promise.resolve().then(()=>this.source.capture(a.signal)).then(p=>{if(!(a.signal.aborted||i.settled))try{kt(p),k(i,{status:"success",snapshot:p,captureMs:Math.max(0,this.now()-s)})}catch(o){k(i,{status:"failure",error:F("capture","invalid_data",!1,o)})}},p=>{a.signal.aborted||i.settled||k(i,{status:"failure",error:F("capture",Ve(p),!1,p)})}).finally(()=>{u.timeoutHandle!==void 0&&clearTimeout(u.timeoutHandle),this.pendingCapture===u&&(this.pendingCapture=void 0)});return Object.assign(u,{controller:a,result:i,settlement:d}),u.timeoutHandle=setTimeout(()=>{i.settled||(a.abort(),k(i,{status:"failure",error:F("capture","timeout",!1)}),this.watchCaptureSettlement(u))},n),this.pendingCapture=u,{turnId:t,generation:r,result:i.promise}}async cancelPendingCaptureAndWait(){if(this.fault!==void 0)throw this.faultedControlError("capture");let t=this.pendingCapture;if(t!==void 0){t.controller.abort(),k(t.result,{status:"cancelled"});try{await this.settleWithin(t.settlement)}catch(r){let n=this.markFault("capture","operation_failed",r);throw new V(n,t.turnId,t.generation)}}}enqueueControl(t){let r=this.controlTail.then(t,t);return this.controlTail=r.catch(()=>{}),r}async applyEnabled(t,r){if(this.fault!==void 0)throw this.faultedControlError(t?"start":"stop");t?await this.startSession(r):await this.stopSession()}async startSession(t){if(this.active)return;let r=this.source;if(r===void 0)throw F("start","not_configured",!1);let n=++this.sessionIdentity,a=new AbortController;this.sessionController=a;let i=Promise.resolve().then(()=>r.start(a.signal)),s={controller:a,settlement:i.then(()=>{})};this.pendingStart=s;try{if(await Ot(i,a.signal),a.signal.aborted||n!==this.sessionIdentity||this.stopping)throw se();this.active=!0}catch(u){throw this.active=!1,Lt(u)||a.signal.aborted?u:F("start",Ve(u),!1,u)}finally{s.settlement.finally(()=>{this.pendingStart===s&&(this.pendingStart=void 0)}).catch(()=>{})}}async stopSession(){this.active=!1,this.abortPendingWork();let t=this.source;if(t!==void 0)try{await this.stopSourceAfterCaptureSettlement(t,!1),this.sessionController=void 0}catch(r){throw this.markFault("stop","operation_failed",r)}}async stopSourceAfterCaptureSettlement(t,r){let n=this.pendingStart?.settlement??Promise.resolve(),a=this.pendingCapture?.settlement??Promise.resolve();try{await this.settleWithin(a.catch(()=>{}))}catch(i){throw r&&await this.settleWithin(Promise.all([n.catch(()=>{}),Promise.resolve().then(()=>t.stop())])).catch(()=>{}),i}await this.settleWithin(Promise.all([n.catch(()=>{}),Promise.resolve().then(()=>t.stop())]).then(()=>{}))}abortPendingWork(){this.sessionController?.abort(),this.pendingStart?.controller.abort();let t=this.pendingCapture;t!==void 0&&(t.controller.abort(),k(t.result,{status:"cancelled"}))}async watchCaptureSettlement(t){try{await this.settleWithin(t.settlement)}catch(r){let n=this.markFault("capture","operation_failed",r);this.onFault?.(n,t.turnId,t.generation)}}settleWithin(t){return new Promise((r,n)=>{let a=setTimeout(()=>{n(new Error("Camera cancellation settlement deadline exceeded"))},this.settlementDeadlineMs);t.then(()=>{clearTimeout(a),r()},i=>{clearTimeout(a),n(i)})})}markFault(t,r,n){return this.fault===void 0&&(this.fault=F(t,r,!0,n)),this.active=!1,this.fault}faultedControlError(t){return F(t,"operation_failed",!0,this.fault)}};function kt(e){if(!(e.data instanceof Uint8Array)||e.data.byteLength===0)throw new Error("Camera snapshot bytes are empty");if(!/^image\/[a-z0-9.+-]+$/i.test(e.mimeType))throw new Error("Camera snapshot MIME is invalid");if(!Number.isInteger(e.width)||e.width<=0)throw new Error("Camera snapshot width is invalid");if(!Number.isInteger(e.height)||e.height<=0)throw new Error("Camera snapshot height is invalid")}function F(e,t,r,n){return new O("Camera operation failed",{role:"camera",operation:e,reason:t,fatal:r,cause:n})}function Ve(e){let t=e instanceof Error?e.name:"";return t==="NotAllowedError"||t==="SecurityError"?"permission_denied":t==="NotFoundError"||t==="NotReadableError"||t==="OverconstrainedError"?"device_unavailable":t==="NotSupportedError"?"unsupported":"operation_failed"}function Mt(){let e,t;return{promise:new Promise((n,a)=>{e=n,t=a}),resolve(n){e(n)},reject(n){t(n)},settled:!1}}function k(e,t){e.settled||(e.settled=!0,e.resolve(t))}function Ot(e,t){return t.aborted?Promise.reject(se()):new Promise((r,n)=>{let a=()=>n(se());t.addEventListener("abort",a,{once:!0}),e.then(r,n).finally(()=>{t.removeEventListener("abort",a)}).catch(()=>{})})}function se(){return new DOMException("Camera operation aborted","AbortError")}function Lt(e){return e instanceof Error&&e.name==="AbortError"}var _=class{constructor(t){this.config=t;this.agentMetadata=C(t.metadata??{}),this.cameraController=new q({...t.transports?.camera!==void 0?{source:t.transports.camera}:{},...t.now!==void 0?{now:t.now}:{},onFault:(r,n)=>{this.reportCameraFault(r,this.envelope("speech",n,this.agentMetadata))}})}config;listeners=new Set;tasks=new Set;pendingAsrControllers=new Map;turnScopes=Fe();cameraController;cameraCaptures=new Map;cameraFaultReported=!1;admissionTail=Promise.resolve();inputControlTail=Promise.resolve();rootController;started=!1;stopping=!1;inputEnabled=!0;inputSessionCounter=0;inputSession;speechGeneration=0;turnCounter=0;skipTts=!1;activeTtsTurn;committedHistory=[];turnTimings=new Map;messages=[];usedTurnIds=new Set;agentMetadata;messageCounter=0;onEvent(t){return this.listeners.add(t),()=>{this.listeners.delete(t)}}async start(){if(this.started)return;this.started=!0,this.rootController=new AbortController,this.inputEnabled&&this.canRunSpeechInput()&&await this.serializeInputControl(()=>this.reconcileInputSession());try{await this.cameraController.startRuntime()}catch(r){this.emit(x(this.envelope("camera",void 0,this.agentMetadata),w(r,{message:"Camera session failed",fatal:!1})))}let t=this.config.greeting;t!==void 0&&t.mode!=="disabled"&&this.track(this.scheduleGreeting(t))}async stop(){this.stopping=!0,this.inputEnabled=!1;let t=this.turnScopes.current();t!==void 0&&this.emitTurnLatency(t),this.rootController?.abort(),this.speechGeneration+=1,this.abortPendingAsr({emitLatency:!1});let r=this.inputSession===void 0?Promise.resolve():this.releaseInputSession(this.inputSession),n=this.cameraController.stopRuntime();this.turnScopes.stop();let a;try{await r}catch(i){a=i}try{await n}catch(i){a??=i}try{await this.config.transports?.output.stop()}catch(i){a=i}try{await this.config.transports?.aec.release()}catch(i){a??=i}if(a!==void 0)throw w(a,{message:"Dialogue runtime stop failed"})}async drain(){for(;this.tasks.size>0;)await Promise.allSettled([...this.tasks])}getMessages(){return this.messages.map(t=>({...t,metadata:C(t.metadata)}))}scheduleGreeting(t){let r=this.reserveTurnId(),n="greeting",a=this.envelope(n,r,this.agentMetadata);return this.beginTurnTiming(r,"greeting"),this.serializeAdmission(async()=>{if(this.rootController?.signal.aborted===!0)return;let i=this.turnScopes.begin({streamId:n,turnId:r});this.track(this.runAssistantTurn(n,r,t.mode==="dynamic"?t.prompt:t.text,a,i,{commitHistory:!1,...t.mode==="static"?{staticReply:t.text}:{},messageMetadata:this.agentMetadata,recordAssistant:!0}))})}async submitText(t,r={}){if(t.trim().length===0)return;this.started||await this.start();let n=this.reserveTurnId(r.turnId),a="manual-text",i=this.effectiveMetadata(r.metadata),s=this.envelope(a,n,i),u={kind:"text",streamId:a,turnId:n,partial:!1,final:!0,metadata:i,text:t};this.beginTurnTiming(n,"text"),await this.serializeAdmission(async()=>{this.speechGeneration+=1,this.abortPendingAsr(),await this.cancelCameraCaptureWithoutBlocking(s),await this.interruptActiveTurn(s,"manual_text"),this.commitMessage(n,"user",t,i),this.emit(Ge(u,"text"));let c=this.turnScopes.begin({streamId:a,turnId:n});this.track(this.runAssistantTurn(a,n,t,s,c,{messageMetadata:i}))})}async setSkipTts(t){if(this.skipTts===t||(this.skipTts=t,!t))return;let r=this.activeTtsTurn;if(!(r===void 0||!this.turnScopes.isCurrent(r.scope))){r.aggregator.clear(),r.worker.clearAndAbort();try{await this.config.transports?.output.flush(),this.activeTtsTurn===r&&this.turnScopes.isCurrent(r.scope)&&await r.playback.close()}catch(n){if(this.activeTtsTurn===r&&this.turnScopes.isCurrent(r.scope)){let a=E(n,{provider:"runtime"});throw this.emit(x(r.base,a)),a}throw E(n,{provider:"runtime"})}}}async setAudioInputEnabled(t){if(this.stopping)throw new h("Dialogue runtime is stopped",{fatal:!0});if(this.inputEnabled===t)return this.inputControlTail;if(this.inputEnabled=t,t||(this.speechGeneration+=1,this.abortPendingAsr({emitLatency:!1}),await this.cancelCameraCaptureWithoutBlocking(this.envelope("camera",void 0,this.agentMetadata)),this.inputSession!==void 0&&(this.inputSession.controller.abort(),this.releaseInputSession(this.inputSession).catch(()=>{}))),!!this.started)return this.serializeInputControl(()=>this.reconcileInputSession())}async setCameraCaptureEnabled(t){if(this.stopping)throw new h("Dialogue runtime is stopped",{fatal:!0});await this.cameraController.setEnabled(t)}serializeInputControl(t){let r=this.inputControlTail.then(t,t);return this.inputControlTail=r.catch(()=>{}),r}async reconcileInputSession(){let t=this.inputSession;if(!this.started||this.stopping||!this.inputEnabled||!this.canRunSpeechInput()){t!==void 0&&await this.releaseInputSession(t);return}t!==void 0&&!t.controller.signal.aborted||(t!==void 0&&await this.releaseInputSession(t),!(this.stopping||!this.inputEnabled||!this.canRunSpeechInput())&&await this.startInputSession())}async startInputSession(){let t=this.config.transports;if(t===void 0||this.config.providers.vad===void 0)return;let r={identity:++this.inputSessionCounter,controller:new AbortController};this.inputSession=r;let n=t.input;try{if(await n.start(),!this.isCurrentInputSession(r)){await this.releaseInputSession(r);return}let a=n.frames(r.controller.signal),i=this.runSpeechLoop(r.controller.signal,a,r.identity);this.track(i),i.finally(()=>{this.isCurrentInputSession(r)&&!r.controller.signal.aborted&&this.releaseInputSession(r).catch(()=>{})}).catch(()=>{})}catch(a){let i=r.controller.signal.aborted||!this.inputEnabled||this.stopping;if(await this.releaseInputSession(r),i)return;throw this.inputEnabled=!1,w(a,{message:"Audio input session failed"})}}releaseInputSession(t){if(t.releasePromise!==void 0)return t.releasePromise;t.controller.abort();let r=this.inputSession===t,n=this.config.transports?.input;return t.releasePromise=(async()=>{try{r&&await n?.stop()}catch(a){throw w(a,{message:"Audio input release failed"})}finally{this.inputSession===t&&(this.inputSession=void 0)}})(),t.releasePromise}isCurrentInputSession(t){return this.inputSession?.identity===t.identity&&!t.controller.signal.aborted&&this.inputEnabled&&!this.stopping}canRunSpeechInput(){return this.config.transports!==void 0&&this.config.providers.vad!==void 0}async runSpeechLoop(t,r,n){let a=this.config.transports,i=this.config.providers.vad;if(a===void 0||i===void 0)return;let s="speech",u,c,d=new AbortController,p=()=>d.abort();t.aborted?p():t.addEventListener("abort",p,{once:!0});let o=d.signal;try{let l=this.nearEndFrames(r,s,o),m=De(l),v=(async()=>{try{for await(let y of i.run(m.vadAudio(),{signal:o})){if(o.aborted)return;y.state==="started"?await this.serializeAdmission(async()=>{if(o.aborted)return;this.speechGeneration+=1,c=this.speechGeneration,u=this.reserveTurnId();let S=this.beginTurnTiming(u,"speech");S.recordStageMetadata("vad",y.metadata),S.markVadStarted(),this.abortPendingAsr(),m.start(u,c);let A=this.envelope(s,u,this.agentMetadata);if(await this.interruptActiveTurn(A,"user_speech"),o.aborted||c!==this.speechGeneration)return;this.emit(G(A,"speech.started"));let P;try{P=await this.cameraController.beginCapture(u,c,this.config.camera?.captureTimeoutMs??1500)}catch(ae){this.reportCameraFaultFromUnknown(ae,A,"Camera capture failed")}P!==void 0&&this.cameraCaptures.set(u,this.settleCameraCapture(P,A))}):y.state==="stopped"&&u!==void 0&&c!==void 0&&(m.stop(),this.emit(G(this.envelope(s,u,this.agentMetadata),"speech.stopped")),u=void 0,c=void 0)}}catch(y){let S=o.aborted||t.aborted;throw p(),this.abortPendingAsr({emitLatency:!S,inputSessionIdentity:n}),S||this.turnScopes.cancel(),y}finally{m.close()}})(),f=(async()=>{try{for await(let y of m.utterances()){if(o.aborted)return;if(!this.isCurrentSpeechGeneration(y.generation))continue;let S=Ut(o);this.pendingAsrControllers.set(S.controller,{streamId:s,turnId:y.turnId,inputSessionIdentity:n});let A=!1;try{A=await this.runAsr(y.audio,s,y.turnId,this.envelope(s,y.turnId,this.agentMetadata),y.generation,S.controller.signal)}finally{this.pendingAsrControllers.delete(S.controller),S.unlink()}!A&&this.isCurrentSpeechGeneration(y.generation)&&!o.aborted&&this.emitTurnLatency({streamId:s,turnId:y.turnId})}}catch(y){let S=o.aborted||t.aborted;throw p(),this.abortPendingAsr({emitLatency:!S,inputSessionIdentity:n}),S||this.turnScopes.cancel(),m.close(),y}})(),b=(await Promise.allSettled([v,f])).find(y=>y.status==="rejected");if(b!==void 0)throw b.reason}catch(l){!B(l)&&!t.aborted&&this.emit(x(this.envelope(s,u,this.agentMetadata),E(l,{provider:"runtime"})))}finally{this.abortPendingAsr({emitLatency:!1,inputSessionIdentity:n}),t.removeEventListener("abort",p)}}async runAsr(t,r,n,a,i,s){let u=!1,c=!1,d=async()=>{c||s.aborted||!this.isCurrentSpeechGeneration(i)||(c=!0,await this.cancelCameraCaptureWithoutBlocking(a))};try{let p=this.turnTimings.get(n),o=!1;p?.markAsrStarted();for await(let l of this.config.providers.asr.run(t,{signal:s})){if(s.aborted||!this.isCurrentSpeechGeneration(i))return!1;let m={...l,streamId:r,turnId:n};if(p?.recordStageMetadata("asr",l.metadata),l.final){if(o)continue;let v=!1;if(await this.serializeAdmission(async()=>{s.aborted||!this.isCurrentSpeechGeneration(i)||(p?.markAsrFinal(),l.text.trim().length>0&&this.commitMessage(n,"user",l.text,this.agentMetadata),this.emit(Ge(m,"speech")),v=l.text.trim().length>0)}),v){let f=await this.cameraCaptures.get(n);await this.serializeAdmission(async()=>{if(s.aborted||!this.isCurrentSpeechGeneration(i))return;let g=this.turnScopes.begin({streamId:r,turnId:n});this.track(this.runAssistantTurn(r,n,l.text,a,g,{messageMetadata:this.agentMetadata,...f!==void 0?{cameraSnapshot:f}:{}})),u=!0})}else await d();o=!0}else o||this.emit(Dt(m,"speech"))}return u||await d(),u}catch(p){return!B(p)&&!s.aborted&&this.isCurrentSpeechGeneration(i)&&this.emit(x(a,E(p,{provider:"asr"}))),await d(),!1}finally{this.cameraCaptures.delete(n)}}async runAssistantTurn(t,r,n,a,i,s={}){let u=i.signal,c=ke(),d=Oe({onStarted:()=>{this.turnScopes.isCurrent(i)&&this.emit(G(a,"playback.started"))},onStopped:()=>{this.turnScopes.isCurrent(i)&&this.emit(G(a,"playback.stopped"))}}),p=!1,o=Me({parentSignal:u,process:async(m,v)=>{try{for await(let f of this.ttsFrames(m,t,r,v.signal)){if(!v.isCurrent()||!this.turnScopes.isCurrent(i))return;let g=this.turnTimings.get(r);g?.recordStageMetadata("tts",f.metadata),g?.markTtsFirstAudio();let b=this.config.transports;if(b===void 0)continue;let y=oe(f);if(g?.markPlaybackStarted(),d.start(),await b.output.enqueue(y),!v.isCurrent()||!this.turnScopes.isCurrent(i)||(await b.aec.pushFarEnd(y),!v.isCurrent()||!this.turnScopes.isCurrent(i)))return}}catch(f){!B(f)&&v.isCurrent()&&this.turnScopes.isCurrent(i)&&(p=!0,o.clearAndAbort(),this.emit(x(a,E(f,{provider:"tts"}))))}}}),l={scope:i,base:a,aggregator:c,worker:o,playback:d};this.activeTtsTurn=l;try{if(!this.turnScopes.isCurrent(i))return;this.emit(G(a,"reply.started"));let m="",v=!1,f=s.staticReply!==void 0?jt(s.staticReply,t,r):this.config.providers.llm.run({messages:this.llmMessages(n,s.commitHistory!==!1,s.cameraSnapshot),streamId:t,turnId:r,metadata:s.messageMetadata??this.agentMetadata},{signal:u});for await(let g of f){if(u.aborted||!this.turnScopes.isCurrent(i))return;if(g.text.length>0){let b=this.turnTimings.get(r);if(b?.recordStageMetadata("llm",g.metadata),b?.markLlmFirstToken(),m+=g.text,this.emit(_e(a,"reply.partial",g.text)),!p&&!this.skipTts)for(let y of c.push(g.text))o.enqueue(y)}if(g.final){if(v=!0,!p&&!this.skipTts)for(let b of c.flush())o.enqueue(b);s.commitHistory!==!1&&this.commitHistory(n,m),s.recordAssistant!==!1&&this.commitMessage(r,"assistant",m,s.messageMetadata??this.agentMetadata),this.emit(_e(a,"reply.final",m));break}}if(!this.turnScopes.isCurrent(i))return;if(!v){c.clear(),o.clearAndAbort();return}if(await o.close(),!this.turnScopes.isCurrent(i))return;d.isActive()&&(await this.config.transports?.output.drain(),this.turnScopes.isCurrent(i)&&await d.close())}catch(m){!B(m)&&!u.aborted&&this.turnScopes.isCurrent(i)&&this.emit(x(a,E(m,{provider:"llm"})))}finally{this.turnScopes.isCurrent(i)&&this.emitTurnLatency(i),c.clear(),o.clearAndAbort(),this.activeTtsTurn===l&&(this.activeTtsTurn=void 0),this.turnScopes.complete(i)}}llmMessages(t,r,n){return[...this.config.systemPrompt!==void 0&&this.config.systemPrompt.length>0?[{role:"system",content:this.config.systemPrompt}]:[],...r&&this.config.history!==void 0?this.committedHistory.flatMap(({user:a,assistant:i})=>[{role:"user",content:a},{role:"assistant",content:i}]):[],{role:"user",content:n===void 0?t:[{type:"text",text:t},{type:"image",data:n.data,mimeType:n.mimeType}]}]}async settleCameraCapture(t,r){let n=await t.result;if(!(!this.isCurrentSpeechGeneration(t.generation)||this.stopping)&&n.status!=="cancelled"){if(n.status==="failure"){this.emit(x(r,n.error));return}return this.emit(Vt(r,n.snapshot,n.captureMs)),n.snapshot}}async cancelCameraCaptureWithoutBlocking(t){try{await this.cameraController.cancelPendingCaptureAndWait()}catch(r){this.reportCameraFaultFromUnknown(r,t,"Camera cancellation failed")}}reportCameraFaultFromUnknown(t,r,n){if(t instanceof V){this.reportCameraFault(t.mediaError,this.envelope("speech",t.turnId,this.agentMetadata));return}let a=t instanceof h?t:w(t,{message:n,fatal:!0});this.reportCameraFault(a,r)}reportCameraFault(t,r){this.cameraFaultReported||(this.cameraFaultReported=!0,this.emit(x(r,t)))}commitHistory(t,r){let n=this.config.history?.maxTurns;if(n===void 0)return;this.committedHistory.push({user:t,assistant:r});let a=this.committedHistory.length-n;a>0&&this.committedHistory.splice(0,a)}ttsFrames(t,r,n,a){return this.config.providers.tts.run({kind:"text",streamId:r,turnId:n,partial:!1,final:!0,metadata:{},text:t},{signal:a})}async*nearEndFrames(t,r,n){let a=0;for await(let i of t){if(n.aborted)return;let s=await this.config.transports?.aec.processNearEnd(i);s!==void 0&&(yield ie(s,{streamId:r,sequence:a++,metadata:{}}))}}emit(t){for(let r of this.listeners)r(t)}track(t){this.tasks.add(t),t.finally(()=>this.tasks.delete(t)).catch(()=>{})}serializeAdmission(t){let r=this.admissionTail.then(t,t);return this.admissionTail=r.catch(()=>{}),r}async interruptActiveTurn(t,r){let n=this.turnScopes.current();if(!(n===void 0||(this.emitTurnLatency(n),this.turnScopes.cancel()===void 0))){try{await this.config.transports?.output.flush()}catch(i){this.emit(x(this.envelope(t.streamId,t.turnId,t.metadata),E(i,{provider:"runtime"})))}this.emit(Gt(t,r))}}abortPendingAsr(t={}){for(let[r,n]of this.pendingAsrControllers)t.inputSessionIdentity!==void 0&&n.inputSessionIdentity!==t.inputSessionIdentity||(t.emitLatency!==!1?this.emitTurnLatency(n):this.turnTimings.delete(n.turnId),r.abort(),this.pendingAsrControllers.delete(r))}isCurrentSpeechGeneration(t){return t===this.speechGeneration}reserveTurnId(t){let r=t??this.generatedTurnId();if(r.trim().length===0)throw new h("turnId must not be empty",{fatal:!0});if(this.usedTurnIds.has(r))throw new h("turnId must be unique within an agent session",{fatal:!0});return this.usedTurnIds.add(r),r}generatedTurnId(){do this.turnCounter+=1;while(this.usedTurnIds.has(`turn-${this.turnCounter}`));return`turn-${this.turnCounter}`}beginTurnTiming(t,r){let n=Le({turnId:t,source:r,...this.config.now!==void 0?{now:this.config.now}:{}});return this.turnTimings.set(t,n),n}emitTurnLatency(t){let n=this.turnTimings.get(t.turnId)?.takeSnapshot();this.turnTimings.delete(t.turnId),n!==void 0&&this.emit(_t(this.envelope(t.streamId,t.turnId,this.agentMetadata),n))}envelope(t,r,n={}){return{streamId:t,...r!==void 0?{turnId:r}:{},partial:!1,final:!1,metadata:{...n}}}effectiveMetadata(t){try{let r=C(t??{});return C({...this.agentMetadata,...r})}catch(r){throw new h("Turn metadata must be JSON-compatible",{fatal:!0,cause:r})}}commitMessage(t,r,n,a){this.messageCounter+=1,this.messages.push({id:`message-${this.messageCounter}`,turnId:t,role:r,content:n,createdAt:(this.config.now??Date.now)(),metadata:C(a)})}};async function*jt(e,t,r){yield{kind:"llm",streamId:t,turnId:r,partial:!1,final:!0,metadata:{},text:e}}function Ge(e,t){return{...Ue(e),type:"transcript.final",partial:!1,final:!0,text:e.text,source:t}}function Dt(e,t){return{...Ue(e),type:"transcript.partial",partial:!0,final:!1,text:e.text,source:t}}function G(e,t){return{...e,type:t,partial:!1,final:!0}}function Vt(e,t,r){return{...e,type:"image.captured",partial:!1,final:!0,image:{mimeType:t.mimeType,width:t.width,height:t.height,sizeBytes:t.data.byteLength,captureMs:r}}}function _e(e,t,r){return{...e,type:t,partial:t==="reply.partial",final:t==="reply.final",text:r}}function Gt(e,t){return{...e,type:"interruption",partial:!1,final:!0,reason:t}}function _t(e,t){return{...e,type:"turn.latency",partial:!1,final:!0,latency:t}}function x(e,t){return{...e,type:"error",partial:!1,final:!0,error:t}}function Ue(e){return{streamId:e.streamId,...e.turnId!==void 0?{turnId:e.turnId}:{},...e.sequence!==void 0?{sequence:e.sequence}:{},partial:e.partial,final:e.final,...e.timestamp!==void 0?{timestamp:e.timestamp}:{},metadata:e.metadata,...e.frameId!==void 0?{frameId:e.frameId}:{}}}function B(e){return e instanceof Error&&e.name==="AbortError"}function Ut(e){let t=new AbortController,r=()=>t.abort();return e.aborted?(r(),{controller:t,unlink(){}}):(e.addEventListener("abort",r,{once:!0}),{controller:t,unlink(){e.removeEventListener("abort",r)}})}async function H(e,t={}){if(ue(e.channels,"channels"),ue(e.sourceSampleRate,"sourceSampleRate"),ue(e.targetSampleRate,"targetSampleRate"),e.sourceSampleRate===e.targetSampleRate)return new z(e);let r=await(t.load??Nt)(),n=Wt(r),a=await n.create(e.channels,e.sourceSampleRate,e.targetSampleRate,{converterType:n.ConverterType.SRC_SINC_FASTEST});return new z(e,a)}var z=class{constructor(t,r){this.converter=r;this.channels=t.channels,this.sourceSampleRate=t.sourceSampleRate,this.targetSampleRate=t.targetSampleRate}converter;channels;sourceSampleRate;targetSampleRate;destroyed=!1;simple(t){return this.assertUsable(t),this.converter?.simple(t)??t}full(t){return this.assertUsable(t),this.converter?.full(t)??t}destroy(){this.destroyed||(this.destroyed=!0,this.converter?.destroy())}assertUsable(t){if(this.destroyed)throw new Error("Resampler has been destroyed");if(t.length%this.channels!==0)throw new Error("Interleaved audio length must be divisible by channels")}};async function Nt(){return import("@alexanderolsen/libsamplerate-js")}function Wt(e){if(typeof e!="object"||e===null)throw new Error("libsamplerate module did not load as an object");let t=e,r=t.default??t;if(typeof r.create!="function"||typeof r.ConverterType?.SRC_SINC_FASTEST!="number")throw new Error("libsamplerate module has an incompatible API");return r}function ue(e,t){if(!Number.isFinite(e)||!Number.isInteger(e)||e<=0)throw new Error(`${t} must be a finite positive integer`)}var qt="https://eva-gateway-ali.dev.autoarkai.com",J={asr:"/v1/audio/transcriptions",llm:"/llm/v1/chat/completions",tts:"/v1/audio/speech"};function K(e){return`${qt}${e}`}function Bt(){return new DOMException("Operation aborted","AbortError")}function zt(e){if(e?.aborted===!0)throw Bt()}function Ne(e){let t=e.trim();if(t.startsWith("data:"))return t.slice(5).trimStart()}async function*$(e,t={}){let{signal:r,isTerminator:n}=t,a=new TextDecoder,i="",s=!1;for await(let c of e){if(zt(r),s)continue;i+=a.decode(c,{stream:!0});let d=i.indexOf(`
+`);for(;d>=0;){let p=i.slice(0,d);i=i.slice(d+1);let o=Ne(p);if(o!==void 0){if(n?.(o)===!0){s=!0,i="";break}yield o}d=i.indexOf(`
+`)}}if(s)return;i+=a.decode();let u=Ne(i);u!==void 0&&n?.(u)!==!0&&(yield u)}function de(e){return e==="[DONE]"}async function*Y(e){let t=e.getReader(),r=!1;try{for(;;){let{value:n,done:a}=await t.read();if(a===!0){r=!0;break}n!==void 0&&(yield n)}}finally{try{r||await t.cancel().catch(()=>{})}finally{t.releaseLock()}}}function We(e){let t=atob(e),r=new Uint8Array(t.length);for(let n=0;n<t.length;n+=1)r[n]=t.charCodeAt(n);return r}function le(e){try{return JSON.parse(e)}catch{return}}function Ht(e){try{return We(e)}catch{return}}function ce(e,t,r){return{kind:"asr",streamId:t.streamId,...t.turnId!==void 0?{turnId:t.turnId}:{},partial:!r,final:r,metadata:{},text:e}}function qe(e,t,r){return{kind:"llm",streamId:t.streamId,...t.turnId!==void 0?{turnId:t.turnId}:{},partial:!r,final:r,metadata:{},text:e}}function Be(e,t,r){return{kind:"tts.audio",streamId:t.streamId,...t.turnId!==void 0?{turnId:t.turnId}:{},partial:!r,final:r,metadata:{},audio:e,sampleRate:t.sampleRate,channels:t.channels}}async function*ze(e,t){let r="",n=!1;for await(let a of e){let i=le(a);i!==void 0&&(i.type==="transcript.text.delta"?(r+=i.delta??"",yield ce(r,t,!1)):i.type==="transcript.text.done"&&(yield ce(i.text??r,t,!0),n=!0))}n||(yield ce(r,t,!0))}async function*He(e,t){for await(let r of e){let n=le(r);if(n===void 0)continue;let a=n.choices?.[0]?.delta?.content;typeof a=="string"&&a.length>0&&(yield qe(a,t,!1))}yield qe("",t,!0)}async function*Je(e,t){let r,n=!1;for await(let a of e){if(n)continue;let i=le(a);if(i!==void 0)if(i.type==="speech.audio.delta"&&typeof i.audio=="string"){let s=Ht(i.audio);if(s===void 0)continue;r!==void 0&&(yield Be(r,t,!1)),r=s}else i.type==="speech.audio.done"&&(n=!0)}r!==void 0&&(yield Be(r,t,!0))}function Q(e){return e!==void 0?e:globalThis.fetch.bind(globalThis)}function X(e){return{Authorization:`Bearer ${e}`}}function Jt(e){return e instanceof DOMException&&e.name==="AbortError"}async function Z(e,t,r,n){let a;try{a=await e(t,r)}catch(i){throw Jt(i)?i:L(i,{provider:n})}if(!a.ok){let i;try{i=await a.text()}catch{i=void 0}throw L(i,{provider:n,statusCode:a.status})}return a}function ee(e,t){let r=e.body;if(r===null)throw L("empty response body",{provider:t});return r}var Kt=16e3,re=1;function $t(){return new DOMException("Operation aborted","AbortError")}function pe(e){if(e?.aborted===!0)throw $t()}function Yt(e){let t=e.reduce((a,i)=>a+i.length,0),r=new Uint8Array(t),n=0;for(let a of e)r.set(a,n),n+=a.length;return r}async function Qt(e,t,r){te(t.sampleRate,"ASR target sampleRate"),te(t.channels??re,"ASR fallback channels");let n=t.createResampler??H,a=[],i,s;for await(let d of e){if(pe(r),te(d.sampleRate,"ASR source sampleRate"),te(d.channels,"ASR source channels"),i===void 0)i=d.sampleRate,s=d.channels;else if(d.sampleRate!==i||d.channels!==s)throw new RangeError("ASR source sampleRate and channels must remain stable within an utterance");er(d.audio,d.channels),a.push(d.audio)}pe(r);let u=Yt(a);if(i===void 0||s===void 0)return{bytes:u,sampleRate:t.sampleRate,channels:t.channels??re};if(i===t.sampleRate)return{bytes:u,sampleRate:t.sampleRate,channels:s};let c=await n({channels:s,sourceSampleRate:i,targetSampleRate:t.sampleRate});try{let d=c.simple(Xt(u));if(d.length%s!==0)throw new RangeError("Resampled audio is not aligned to its channel count");return{bytes:Zt(d),sampleRate:t.sampleRate,channels:s}}finally{c.destroy()}}function Xt(e){let t=new Float32Array(e.byteLength/2),r=new DataView(e.buffer,e.byteOffset,e.byteLength);for(let n=0;n<t.length;n+=1)t[n]=r.getInt16(n*2,!0)/32768;return t}function Zt(e){let t=new Uint8Array(e.length*2),r=new DataView(t.buffer);for(let n=0;n<e.length;n+=1){let a=Math.max(-1,Math.min(1,e[n])),i=a<0?Math.round(a*32768):Math.round(a*32767);r.setInt16(n*2,i,!0)}return t}function er(e,t){if(e.byteLength%2!==0)throw new RangeError("ASR PCM16 frame must contain an even number of bytes");if(e.byteLength/2%t!==0)throw new RangeError("ASR PCM16 frame must align to its channel count")}function te(e,t){if(!Number.isFinite(e)||!Number.isInteger(e)||e<=0)throw new RangeError(`${t} must be a finite positive integer`)}function tr(e){return Symbol.asyncIterator in Object(e)}async function rr(e,t){if(tr(e)){let r="",n="tts",a;for await(let i of e)pe(t),r+=i.text,n=i.streamId,a=i.turnId;return a!==void 0?{text:r,streamId:n,turnId:a}:{text:r,streamId:n}}return e.turnId!==void 0?{text:e.text,streamId:e.streamId,turnId:e.turnId}:{text:e.text,streamId:e.streamId}}function nr(e){return e.map(t=>{if(Array.isArray(t.content)&&t.content.length===0)throw new h("Gateway LLM content parts must not be empty",{fatal:!0});let r=typeof t.content=="string"?t.content:t.content.map(a=>{if(a.type==="text")return{type:"text",text:a.text};if(a.data.byteLength===0||!/^image\/[a-z0-9.+-]+$/i.test(a.mimeType))throw new h("Gateway LLM image content is invalid",{fatal:!0});return{type:"image_url",image_url:{url:`data:${a.mimeType};base64,${ar(a.data)}`}}}),n={role:t.role,content:r};return t.name!==void 0&&(n.name=t.name),t.toolCallId!==void 0&&(n.tool_call_id=t.toolCallId),n})}function ar(e){let r="";for(let n=0;n<e.length;n+=32768)r+=String.fromCharCode(...e.subarray(n,n+32768));return btoa(r)}function me(e){let t=Q(e.fetch),r=X(e.apiKey);return{run(n,a){return(async function*(){let s=a?.signal,u;try{u=await Qt(n,e,s)}catch(f){throw f instanceof DOMException&&f.name==="AbortError"?f:E(f,{provider:"gateway-asr",message:"Gateway ASR audio preprocessing failed"})}let{bytes:c,sampleRate:d,channels:p}=u,o=new FormData;o.append("model",e.model),o.append("stream","true"),o.append("audio_format","pcm"),o.append("sample_rate",String(d)),o.append("channels",String(p)),e.hotwords!==void 0&&o.append("hotwords",e.hotwords),o.append("file",new Blob([c],{type:"application/octet-stream"}),"audio.pcm");let l={method:"POST",headers:r,body:o};s!==void 0&&(l.signal=s);let m=await Z(t,K(J.asr),l,"asr"),v=$(Y(ee(m,"asr")),{...s!==void 0?{signal:s}:{},isTerminator:de});yield*ze(v,{streamId:"speech"})})()}}}function fe(e){let t=Q(e.fetch),n={...X(e.apiKey),"Content-Type":"application/json"};return{run(a,i){return(async function*(){let u=i?.signal,c={model:e.model,stream:!0,messages:nr(a.messages)};e.temperature!==void 0&&(c.temperature=e.temperature),e.maxTokens!==void 0&&(c.max_tokens=e.maxTokens),e.topP!==void 0&&(c.top_p=e.topP);let d={method:"POST",headers:n,body:JSON.stringify(c)};u!==void 0&&(d.signal=u);let p=await Z(t,K(J.llm),d,"llm"),o=$(Y(ee(p,"llm")),{...u!==void 0?{signal:u}:{},isTerminator:de}),l=a.turnId!==void 0?{streamId:a.streamId,turnId:a.turnId}:{streamId:a.streamId};yield*He(o,l)})()}}}function he(e){let t=Q(e.fetch),n={...X(e.apiKey),"Content-Type":"application/json"};return{run(a,i){return(async function*(){let u=i?.signal,{text:c,streamId:d,turnId:p}=await rr(a,u),o=e.sampleRate??Kt,l={model:e.model,input:c,response_format:"pcm",stream_format:"sse",sample_rate:o};e.voice!==void 0&&(l.voice=e.voice),e.speed!==void 0&&(l.speed=e.speed),e.pitchRate!==void 0&&(l.pitch_rate=e.pitchRate);let m={method:"POST",headers:n,body:JSON.stringify(l)};u!==void 0&&(m.signal=u);let v=await Z(t,K(J.tts),m,"tts"),f=$(Y(ee(v,"tts")),{...u!==void 0?{signal:u}:{}});yield*Je(f,p!==void 0?{streamId:d,turnId:p,sampleRate:o,channels:re}:{streamId:d,sampleRate:o,channels:re})})()}}}import*as M from"onnxruntime-web";var ir=new URL("./assets/silero_vad_v6.onnx",import.meta.url).href;async function Ke(e={},t){let r=e.modelUrl??ir,n=await(e.modelFetcher??or)(r,t),a=await M.InferenceSession.create(n);return{async run(i){let s=await a.run({input:new M.Tensor("float32",i.input,[1,i.input.length]),state:new M.Tensor("float32",i.state,[2,1,128]),sr:new M.Tensor("int64",BigInt64Array.from([BigInt(i.sampleRate)]),[])}),u=s.output,c=s.stateN;if(u===void 0||c===void 0||u.type!=="float32"||c.type!=="float32"||!(u.data instanceof Float32Array)||!(c.data instanceof Float32Array)||u.data.length!==1||!sr(c.dims,[2,1,128]))throw new Error("Silero VAD v6 returned an invalid result");return ge({speechProbability:u.data[0],state:Float32Array.from(c.data)})}}}function ge(e){if(!Number.isFinite(e.speechProbability)||e.speechProbability<0||e.speechProbability>1||e.state.length!==256||!e.state.every(Number.isFinite))throw new Error("Silero VAD v6 returned an invalid result");return e}async function or(e,t){let r=await fetch(e,t!==void 0?{signal:t}:{});if(!r.ok)throw new Error("Silero VAD model fetch failed");return r.arrayBuffer()}function sr(e,t){return e.length===t.length&&e.every((r,n)=>r===t[n])}var $e=16e3,ye=512,ne=64;function be(e={}){return new Se(e)}var Se=class{constructor(t){this.options=t}options;run(t,r){return this.runFrames(t,r)}async*runFrames(t,r){let n=r?.signal,a=this.options.positiveSpeechThreshold??.5,i=this.options.negativeSpeechThreshold??.35,s=Math.max(1,Math.ceil((this.options.silenceThresholdMs??200)/32)),u=!1,c=0,d=new Float32Array(256),p=new Float32Array(ne),o,l,m,v=[];try{if(U(n))return;let f=await Ye(this.options.createSession!==void 0?this.options.createSession():Ke(this.options,n),n);for await(let g of t){if(U(n))return;if(o=g,l!==void 0&&g.sampleRate!==l)throw new RangeError("VAD source sampleRate cannot change within a run");l??=g.sampleRate,m??=await H({channels:1,sourceSampleRate:l,targetSampleRate:$e});let b=ur(g);for(v.push(...m.full(b));v.length>=ye;){let y=Float32Array.from(v.splice(0,ye)),S=new Float32Array(ne+ye);S.set(p),S.set(y,ne);let A=ge(await Ye(f.run({input:S,state:d,sampleRate:$e}),n));if(U(n))return;d=Float32Array.from(A.state),p=S.slice(S.length-ne),A.speechProbability>=a?(c=0,u||(u=!0,yield ve(g,"started",A.speechProbability))):u&&A.speechProbability<i?(c+=1,c>=s&&(u=!1,c=0,yield ve(g,"stopped",A.speechProbability))):u&&(c=0)}}u&&!U(n)&&o!==void 0&&(yield ve(o,"stopped"))}catch(f){if(U(n))return;throw E(f,{provider:"silero-vad"})}finally{m?.destroy()}}};function U(e){return e?.aborted===!0}function Ye(e,t){return t===void 0?e:t.aborted?Promise.reject(Qe()):new Promise((r,n)=>{let a=()=>{t.removeEventListener("abort",a),n(Qe())};t.addEventListener("abort",a,{once:!0}),e.then(i=>{t.removeEventListener("abort",a),r(i)},i=>{t.removeEventListener("abort",a),n(i)})})}function Qe(){let e=new Error("Operation aborted");return e.name="AbortError",e}function ur(e){if(!Number.isInteger(e.channels)||e.channels<=0)throw new RangeError("Audio frame channels must be positive");let t=e.channels*2;if(e.audio.byteLength%t!==0)throw new RangeError("Audio frame PCM must align to its channel count");let r=e.audio.byteLength/t,n=new Float32Array(r),a=new DataView(e.audio.buffer,e.audio.byteOffset,e.audio.byteLength);for(let i=0;i<r;i+=1){let s=0;for(let u=0;u<e.channels;u+=1)s+=a.getInt16((i*e.channels+u)*2,!0)/32768;n[i]=s/e.channels}return n}function ve(e,t,r){return{kind:"vad",streamId:e.streamId,...e.turnId!==void 0?{turnId:e.turnId}:{},partial:t==="started",final:t==="stopped",metadata:e.metadata,state:t,...r!==void 0?{confidence:r}:{}}}function Ae(e){return e.fetch!==void 0?{fetch:e.fetch}:{}}function Xe(e,t){return me({apiKey:t.apiKey,model:e.model,sampleRate:e.sampleRate,...Ae(t)})}function Ze(e,t){return fe({apiKey:t.apiKey,model:e.model,...e.temperature!==void 0?{temperature:e.temperature}:{},...e.maxTokens!==void 0?{maxTokens:e.maxTokens}:{},...Ae(t)})}function et(e,t){return he({apiKey:t.apiKey,model:e.model,...e.voice!==void 0?{voice:e.voice}:{},...e.speakingRate!==void 0?{speed:e.speakingRate}:{},...e.sampleRate!==void 0?{sampleRate:e.sampleRate}:{},...e.pitch!==void 0?{pitchRate:e.pitch}:{},...Ae(t)})}function tt(e,t){if(e===void 0)return;if(e.sensitivity!==void 0&&!(e.sensitivity>0&&e.sensitivity<=1))throw new h("VAD sensitivity must be within (0, 1]",{fatal:!0});if(e.silenceThresholdMs!==void 0&&(!Number.isFinite(e.silenceThresholdMs)||e.silenceThresholdMs<=0))throw new h("VAD silenceThresholdMs must be finite and greater than 0",{fatal:!0});let r=e.sensitivity??.5;return be({positiveSpeechThreshold:r,negativeSpeechThreshold:Math.max(0,r-.15),...e.silenceThresholdMs!==void 0?{silenceThresholdMs:e.silenceThresholdMs}:{},...t.createSileroSession!==void 0?{createSession:t.createSileroSession}:{}})}var dr=10,cr=1500,lr="\u8BF7\u7528\u4E00\u53E5\u7B80\u77ED\u3001\u81EA\u7136\u7684\u8BDD\u5411\u7528\u6237\u6253\u62DB\u547C\u3002";function Ee(e={}){return{create(t){let r={apiKey:t.apiKey,...e.fetch!==void 0?{fetch:e.fetch}:{},...e.createSileroSession!==void 0?{createSileroSession:e.createSileroSession}:{}},n=tt(t.vad,r);return{asr:Xe(t.asr,r),llm:Ze(t.llm,r),tts:et(t.tts,r),...n!==void 0?{vad:n}:{}}}}}function we(e,t){rt(e.asr.sampleRate,"ASR sampleRate"),e.tts.sampleRate!==void 0&&rt(e.tts.sampleRate,"TTS sampleRate");let r=pr(e.camera?.captureTimeoutMs),n=e.transports?.input!==void 0,a={apiKey:e.apiKey,asr:e.asr,tts:e.tts,llm:e.llm};e.vad!==void 0&&(a.vad=e.vad);let i=t.create(a);if(n&&i.vad===void 0)throw new h("Audio input requires a VAD provider",{fatal:!0});let s={systemPrompt:e.systemPrompt??"",greeting:fr(e.greeting),metadata:mr(e.metadata),camera:{captureTimeoutMs:r},providers:i};return e.history!==void 0&&(s.history={maxTurns:hr(e.history.maxTurns)}),e.transports!==void 0&&(s.transports=e.transports),s}function pr(e){let t=e??cr;if(!Number.isFinite(t)||!Number.isInteger(t)||t<=0)throw new h("Camera captureTimeoutMs must be a finite positive integer",{fatal:!0});return t}function mr(e){try{return C(e??{})}catch(t){throw new h("Agent metadata must be JSON-compatible",{fatal:!0,cause:t})}}function fr(e){if(e===void 0||e.mode==="disabled")return{mode:"disabled"};if(e.mode==="static"){if(e.text.trim().length===0)throw new h("Static greeting text must not be empty",{fatal:!0});return{mode:"static",text:e.text}}return{mode:"dynamic",prompt:e.prompt===void 0||e.prompt.trim().length===0?lr:e.prompt}}function hr(e){let t=e??dr;if(!Number.isInteger(t)||t<=0)throw new h("History maxTurns must be a positive integer",{fatal:!0});return t}function rt(e,t){if(!Number.isFinite(e)||!Number.isInteger(e)||e<=0)throw new h(`${t} must be a finite positive integer`,{fatal:!0})}function nt(e,t,r){return{...T(e),type:"transcript.final",final:!0,partial:!1,text:t,source:r}}function at(e,t){return{...T(e),type:"transcript.partial",final:!1,partial:!0,text:t,source:"speech"}}function it(e){return{...T(e),type:"speech.started",partial:!1,final:!0}}function ot(e,t){return{...T(e),type:"image.captured",partial:!1,final:!0,image:{...t}}}function st(e){return{...T(e),type:"speech.stopped",partial:!1,final:!0}}function ut(e,t){return{...T(e),type:"interruption",partial:!1,final:!0,reason:t}}function dt(e){return{...T(e),type:"reply.started",partial:!1,final:!0}}function ct(e,t){return{...T(e),type:"reply.partial",partial:!0,final:!1,text:t}}function lt(e,t){return{...T(e),type:"reply.final",partial:!1,final:!0,text:t}}function pt(e){return{...T(e),type:"playback.started",partial:!1,final:!0}}function mt(e){return{...T(e),type:"playback.stopped",partial:!1,final:!0}}function ft(e,t){return{...T(e),type:"turn.latency",partial:!1,final:!0,latency:t}}function ht(e,t){return{...e,type:"error",partial:!1,final:!0,error:gr(t)}}function T(e){if(e.turnId===void 0)throw new h("Runtime event is missing turn identity",{fatal:!0});return{...e,turnId:e.turnId}}function gr(e){return{message:e.message,fatal:e.fatal,source:e.source,...e.provider!==void 0?{provider:e.provider}:{},...e.statusCode!==void 0?{statusCode:e.statusCode}:{},...e.role!==void 0?{role:e.role}:{},...e.operation!==void 0?{operation:e.operation}:{},...e.reason!==void 0?{reason:e.reason}:{}}}function Ce(e){let t=new Set,r=new Map,n=e.onEvent(a=>{let i=vr(a,yr(r,a.streamId));if(i!==void 0)for(let s of t)s(i)});return{onEvent(a){return t.add(a),()=>{t.delete(a)}},close(){n(),t.clear()}}}function yr(e,t){let r=e.get(t)??0;return e.set(t,r+1),r}function vr(e,t){let r=Sr(e,t);switch(e.type){case"speech.started":return it(r);case"image.captured":return ot(r,e.image);case"speech.stopped":return st(r);case"transcript.partial":return at(r,e.text);case"transcript.final":return nt(r,e.text,e.source);case"interruption":return ut(r,e.reason);case"reply.started":return dt(r);case"reply.partial":return ct(r,e.text);case"reply.final":return lt(r,e.text);case"playback.started":return pt(r);case"playback.stopped":return mt(r);case"turn.latency":return ft(r,e.latency);case"error":return ht(r,e.error);default:return}}function Sr(e,t){return{streamId:e.streamId,...e.turnId!==void 0?{turnId:e.turnId}:{},sequence:t,partial:e.partial,final:e.final,...e.timestamp!==void 0?{timestamp:e.timestamp}:{},metadata:br(e.metadata),...e.frameId!==void 0?{frameId:e.frameId}:{}}}var gt=/(api.?key|authorization|headers?|raw|body|sse|pcm|provider.?object|secret|token|credential|password|cookies?)/i,I=Symbol("unsafe-metadata");function br(e){let t={};for(let[r,n]of Object.entries(e)){if(gt.test(r))continue;let a=Te(n,new Set);a!==I&&(t[r]=a)}return t}function Te(e,t){if(e===null||typeof e=="string"||typeof e=="boolean")return e;if(typeof e=="number")return Number.isFinite(e)?e:I;if(typeof e!="object"||t.has(e))return I;t.add(e);try{if(Array.isArray(e)){let n=[];for(let a of e){let i=Te(a,t);if(i===I)return I;n.push(i)}return n}if(Object.getPrototypeOf(e)!==Object.prototype&&Object.getPrototypeOf(e)!==null)return I;let r={};for(let[n,a]of Object.entries(e)){if(gt.test(n))continue;let i=Te(a,t);if(i===I)return I;r[n]=i}return r}finally{t.delete(e)}}function yt(e){return Ar(e,Ee())}function Ar(e,t){let r=we(e,t);return Er(new _(r))}function Er(e){let t=Ce(e),r="created",n=!1,a,i,s=new Set,u=(d,p,o=!0)=>{let l;return l=(async()=>{try{if(await d(),n)throw R("Agent media control was cancelled by stop")}catch(m){throw n?R("Agent media control was cancelled by stop"):o&&m instanceof h?m:new h(p,{cause:m})}finally{s.delete(l)}})(),s.add(l),l};return{start(){if(n||r==="stopped")return Promise.reject(R("Agent is stopped"));if(a!==void 0)return a;let d=(async()=>{try{if(await e.start(),n)throw R("Agent start was cancelled by stop");r="running"}catch(p){if(n)throw R("Agent start was cancelled by stop");try{await e.stop()}catch{}throw r="created",a=void 0,p instanceof h?p:w(p,{message:"Agent start failed"})}})();return a=d,d},submitText(d,p){if(n||r!=="running")return Promise.reject(R(r==="created"?"Agent has not started":"Agent is stopped"));let o;try{o=p===void 0?void 0:wr(p)}catch(l){return Promise.reject(w(l,{message:"Turn metadata must be JSON-compatible"}))}return e.submitText(d,o).catch(l=>{throw w(l,{message:"Agent text submission failed"})})},setAudioInputEnabled(d){return n||r==="stopped"?Promise.reject(R("Agent is stopped")):u(()=>e.setAudioInputEnabled(d),"Agent audio input update failed",!1)},setCameraCaptureEnabled(d){return n||r==="stopped"?Promise.reject(R("Agent is stopped")):u(()=>e.setCameraCaptureEnabled(d),"Agent camera capture update failed")},setTtsEnabled(d){return n||r==="stopped"?Promise.reject(R("Agent is stopped")):u(()=>e.setSkipTts(!d),"Agent TTS update failed")},getMessages(){return e.getMessages()},onEvent(d){if(n||r==="stopped")throw R(r==="stopped"?"Agent is stopped":"Agent is stopping");let p=t.onEvent(d),o=!0;return()=>{o&&(o=!1,p())}},stop(){if(i!==void 0)return i;n=!0;let d=a,p=[...s],o=(async()=>{let l;try{try{await e.stop()}catch(m){l=m}if(await Promise.allSettled([...d===void 0?[]:[d],...p]),l!==void 0)throw w(l,{message:"Agent stop failed"})}finally{t.close(),r="stopped"}})();return i=o,o}}}function wr(e){return{...e.turnId!==void 0?{turnId:e.turnId}:{},...e.metadata!==void 0?{metadata:C(e.metadata)}:{}}}function R(e){return new h(e,{fatal:!0})}export{h as EvaSdkError,yt as createEvaVoiceDialogueAgent};