@xmov/avatar 2.4.0 → 2.5.0-alpha.1

package/package-dist/agent/index.d.ts

@@ -0,0 +1,42 @@
+import type { AvatarStatus, RenderState, Layout, WalkConfig, ICapabilityCheckReport } from "../public-types.js";
+import type { XingyunAvatarAgentOptions, XingyunAvatarAgentInitParams, AgentASRStartOptions, AgentState, AgentASRState } from "./public-types.js";
+export * from "./public-types.js";
+export default class XingyunAvatarAgent {
+    static checkCapability(audioOnlyMode?: boolean, hardwareAcceleration?: string): Promise<ICapabilityCheckReport>;
+    constructor(options: XingyunAvatarAgentOptions);
+    init: (params?: XingyunAvatarAgentInitParams) => Promise<void>;
+    destroy: (reason?: string) => Promise<void>;
+    ask: (text: string) => Promise<void>;
+    speak: (ssml: string, is_start?: boolean, is_end?: boolean) => string | undefined;
+    /** @deprecated Use speak instead. */
+    speakByE2E: (text: string) => Promise<void>;
+    startASR: (options?: AgentASRStartOptions) => Promise<void>;
+    stopASR: () => Promise<void>;
+    interrupt: () => number;
+    interruptConversation: (reason?: string) => Promise<void>;
+    getAgentState: () => AgentState;
+    /** @deprecated Use getAgentState instead. */
+    getState: () => AgentState;
+    getASRState: () => AgentASRState;
+    switchInvisibleMode: () => Promise<void>;
+    getStatus: () => AvatarStatus | -1;
+    getRenderState: () => RenderState;
+    getSessionId: () => string | undefined;
+    idle: () => void;
+    listen: () => void;
+    think: () => void;
+    interactiveidle: () => void;
+    setVolume: (volume: number) => void;
+    changeAvatarVisible: (visible: boolean) => void;
+    changeLayout: (layout: Layout) => void;
+    changeWalkConfig: (config: WalkConfig) => void;
+    showDebugInfo: () => void;
+    hideDebugInfo: () => void;
+    restartWebGL: () => boolean;
+    /** @deprecated Retained for compatibility; migration will be announced separately. */
+    startRecoder: () => Promise<void>;
+    /** @deprecated Retained for compatibility; migration will be announced separately. */
+    stopRecoder: () => void;
+    /** @deprecated Starts a browser download; does not return a Blob. */
+    exportVideo: () => void;
+}

package/{dist/agent/types.d.ts → package-dist/agent/public-types.d.ts}

@@ -1,116 +1,115 @@
-import type { IAvatarOptions, IInitParams, INetworkInfo, RenderState } from "../types";
-export type AgentState = "idle" | "initializing" | "ready" | "running" | "reconnecting" | "stopped" | "destroyed" | "failed";
-export type AgentSocketState = "idle" | "connecting" | "open" | "closed" | "error";
-export type AgentASRState = "idle" | "requesting-permission" | "starting" | "listening" | "stopping" | "failed";
-export type AgentConversationState = "idle" | "asking" | "speaking-directly" | "speaking" | "completed" | "interrupted" | "failed";
-export type AgentErrorDomain = "permission" | "network" | "asr" | "brain" | "ttsa" | "quota" | "sdk";
-export interface AgentAudioOptions {
-    chunkMs?: number;
-    inputEnabled?: boolean;
-    /**
-     * @deprecated Session 返回 features.speech_frontend.enable_aec 后，以服务端返回值为准。
-     */
-    echoCancellationEnabled?: boolean;
-}
-export interface AgentReconnectOptions {
-    enabled?: boolean;
-    maxAttempts?: number;
-    initialDelayMs?: number;
-    maxDelayMs?: number;
-}
-export interface AgentError {
-    code: string;
-    message: string;
-    domain: AgentErrorDomain;
-    retryable?: boolean;
-    cause?: unknown;
-}
-export interface AgentASRResult {
-    text: string;
-    isFinal: boolean;
-    raw?: unknown;
-}
-export interface AgentASRStartOptions {
-    /**
-     * Optional caller-owned audio stream used instead of requesting the physical microphone.
-     * The Agent clones one live audio track and never stops the caller-owned stream.
-     */
-    inputStream?: MediaStream;
-}
-export interface AgentLLMUsage {
-    promptTokens: number;
-    completionTokens: number;
-    totalTokens: number;
-    cachedTokens: number;
-}
-export interface AgentLLMResponse {
-    event: "chunk" | "done";
-    text?: string;
-    isFirst?: boolean;
-    usage?: AgentLLMUsage;
-    raw?: unknown;
-}
-export interface AgentSemanticJudgeResult {
-    query: string;
-    meaningful: boolean;
-    action: string;
-    raw?: unknown;
-}
-export interface AgentSpeakStateEvent {
-    state: string;
-    clientSpeakId?: string | number;
-    raw?: unknown;
-}
-export interface AgentConversationEvent {
-    state: AgentConversationState;
-    text?: string;
-    raw?: unknown;
-}
-export interface XingyunAvatarAgentCallbacks {
-    onAgentStateChange?: (state: AgentState) => void;
-    onSocketStateChange?: (state: AgentSocketState) => void;
-    onASRStateChange?: (state: AgentASRState) => void;
-    onASRResult?: (result: AgentASRResult) => void;
-    onLLMResponse?: (response: AgentLLMResponse) => void;
-    onSpeakStateChange?: (event: AgentSpeakStateEvent) => void;
-    onConversationChange?: (event: AgentConversationEvent) => void;
-    onSemanticJudgeResult?: (result: AgentSemanticJudgeResult) => void;
-    onRenderChange?: (state: RenderState) => void;
-    onNetworkInfo?: (info: INetworkInfo) => void;
-    onError?: (error: AgentError) => void;
-}
-export interface AgentBrainConfig extends Record<string, unknown> {
-    extra_body?: Record<string, unknown>;
-}
-export interface AgentFeatures extends Record<string, unknown> {
-    auto_send_asr_to_llm?: boolean;
-    anti_interference?: {
-        semantic_judge_enabled?: boolean;
-    };
-    speech_frontend?: {
-        enabled?: boolean;
-        enable_aec?: boolean;
-        enable_speech_separation?: boolean;
-    };
-    vad_merge_mode?: boolean;
-    volume_and_repetition_text_detection?: boolean;
-}
-export interface XingyunAvatarAgentOptions extends IAvatarOptions {
-    asr_id?: number;
-    asr_config?: Record<string, unknown>;
-    tts_config?: Record<string, unknown>;
-    features?: AgentFeatures;
-    extras?: Record<string, unknown>;
-    llm_id?: number;
-    brain_config?: AgentBrainConfig;
-    session_speak_req_id?: 0 | 1;
-    e2eServer?: string;
-    authToken?: string;
-    audio?: AgentAudioOptions;
-    reconnect?: AgentReconnectOptions;
-    agentCallbacks?: XingyunAvatarAgentCallbacks;
-    webSocketCtor?: typeof WebSocket;
-}
-export type XingyunAvatarAgentInitParams = Omit<IInitParams, "onDownloadProgress"> & {
-    onDownloadProgress?: IInitParams["onDownloadProgress"];
-};
+import type { IAvatarOptions, IInitParams, INetworkInfo, RenderState } from "../public-types.js";
+export type AgentState = "idle" | "initializing" | "ready" | "running" | "reconnecting" | "stopped" | "destroyed" | "failed";
+export type AgentSocketState = "idle" | "connecting" | "open" | "closed" | "error";
+export type AgentASRState = "idle" | "requesting-permission" | "starting" | "listening" | "stopping" | "failed";
+export type AgentConversationState = "idle" | "asking" | "speaking-directly" | "speaking" | "completed" | "interrupted" | "failed";
+export type AgentErrorDomain = "permission" | "network" | "asr" | "brain" | "ttsa" | "quota" | "sdk";
+export interface AgentAudioOptions {
+    chunkMs?: number;
+    inputEnabled?: boolean;
+    /**
+     * @deprecated Session 返回 features.speech_frontend.enable_aec 后，以服务端返回值为准。
+     */
+    echoCancellationEnabled?: boolean;
+}
+export interface AgentReconnectOptions {
+    enabled?: boolean;
+    maxAttempts?: number;
+    initialDelayMs?: number;
+    maxDelayMs?: number;
+}
+export interface AgentError {
+    code: string;
+    message: string;
+    domain: AgentErrorDomain;
+    retryable?: boolean;
+    cause?: unknown;
+}
+export interface AgentASRResult {
+    text: string;
+    isFinal: boolean;
+    raw?: unknown;
+}
+export interface AgentASRStartOptions {
+    /**
+     * Optional caller-owned audio stream used instead of requesting the physical microphone.
+     * The Agent clones one live audio track and never stops the caller-owned stream.
+     */
+    inputStream?: MediaStream;
+}
+export interface AgentLLMUsage {
+    promptTokens: number;
+    completionTokens: number;
+    totalTokens: number;
+    cachedTokens: number;
+}
+export interface AgentLLMResponse {
+    event: "chunk" | "done";
+    text?: string;
+    isFirst?: boolean;
+    usage?: AgentLLMUsage;
+    raw?: unknown;
+}
+export interface AgentSemanticJudgeResult {
+    query: string;
+    meaningful: boolean;
+    action: string;
+    raw?: unknown;
+}
+export interface AgentSpeakStateEvent {
+    state: string;
+    clientSpeakId?: string | number;
+    raw?: unknown;
+}
+export interface AgentConversationEvent {
+    state: AgentConversationState;
+    text?: string;
+    raw?: unknown;
+}
+export interface XingyunAvatarAgentCallbacks {
+    onAgentStateChange?: (state: AgentState) => void;
+    onSocketStateChange?: (state: AgentSocketState) => void;
+    onASRStateChange?: (state: AgentASRState) => void;
+    onASRResult?: (result: AgentASRResult) => void;
+    onLLMResponse?: (response: AgentLLMResponse) => void;
+    onSpeakStateChange?: (event: AgentSpeakStateEvent) => void;
+    onConversationChange?: (event: AgentConversationEvent) => void;
+    onSemanticJudgeResult?: (result: AgentSemanticJudgeResult) => void;
+    onRenderChange?: (state: RenderState) => void;
+    onNetworkInfo?: (info: INetworkInfo) => void;
+    onError?: (error: AgentError) => void;
+}
+export interface AgentBrainConfig extends Record<string, unknown> {
+    extra_body?: Record<string, unknown>;
+}
+export interface AgentFeatures extends Record<string, unknown> {
+    auto_send_asr_to_llm?: boolean;
+    anti_interference?: {
+        semantic_judge_enabled?: boolean;
+    };
+    speech_frontend?: {
+        enabled?: boolean;
+        enable_aec?: boolean;
+        enable_speech_separation?: boolean;
+    };
+    vad_merge_mode?: boolean;
+    volume_and_repetition_text_detection?: boolean;
+}
+export interface XingyunAvatarAgentOptions extends IAvatarOptions {
+    asr_id?: number;
+    asr_config?: Record<string, unknown>;
+    tts_config?: Record<string, unknown>;
+    features?: AgentFeatures;
+    extras?: Record<string, unknown>;
+    llm_id?: number;
+    brain_config?: AgentBrainConfig;
+    session_speak_req_id?: 0 | 1;
+    e2eServer?: string;
+    authToken?: string;
+    audio?: AgentAudioOptions;
+    reconnect?: AgentReconnectOptions;
+    agentCallbacks?: XingyunAvatarAgentCallbacks;
+}
+export type XingyunAvatarAgentInitParams = Omit<IInitParams, "onDownloadProgress"> & {
+    onDownloadProgress?: IInitParams["onDownloadProgress"];
+};