@xmov/avatar 2.1.0 → 2.1.1

package/src/agent/types.ts

@@ -1,197 +0,0 @@
-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 type AgentServiceConfig = Record<string, unknown>;
-
-export interface AgentBrainConfig extends AgentServiceConfig {
-  extra_body?: Record<string, unknown>;
-}
-
-export interface AgentFeatures extends AgentServiceConfig {
-  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?: AgentServiceConfig;
-  tts_config?: AgentServiceConfig;
-  features?: AgentFeatures;
-  extras?: AgentServiceConfig;
-  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"];
-};
-
-export interface AgentSessionInfo {
-  ws_url?: string;
-  e2e_token?: string;
-  monitor_ws_url?: string;
-}
-
-export interface AgentClientMessage {
-  type: string;
-  message?: unknown;
-  [key: string]: unknown;
-}
-
-export interface AgentServerEvent {
-  type: string;
-  domain?: string;
-  code?: string | number;
-  event?: string;
-  message?: unknown;
-  text?: string;
-  is_first?: boolean;
-  is_final?: boolean;
-  usage?: {
-    prompt_tokens?: number;
-    completion_tokens?: number;
-    total_tokens?: number;
-    cached_tokens?: number;
-  };
-  [key: string]: unknown;
-}
-
-export interface AgentAudioMetadata {
-  format: "webm_opus";
-  mimeType: "audio/webm;codecs=opus";
-  /** Fixed output track sample rate passed to the MediaRecorder pipeline. */
-  sampleRate: number;
-  /** Fixed output track channel count passed to the MediaRecorder pipeline. */
-  channels: number;
-  chunkMs: number;
-}
-
-export const DEFAULT_AGENT_AUDIO: AgentAudioMetadata = {
-  format: "webm_opus",
-  mimeType: "audio/webm;codecs=opus",
-  sampleRate: 16000,
-  channels: 1,
-  chunkMs: 100,
-};