@xmov/avatar 2.5.0-alpha.6 → 2.5.0-alpha.8

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

@@ -1,5 +1,5 @@
 import type { AvatarStatus, RenderState, Layout, WalkConfig, ICapabilityCheckReport } from "../public-types.js";
-import type { XingyunAvatarAgentOptions, XingyunAvatarAgentInitParams, AgentASRStartOptions, AgentState, AgentASRState } from "./public-types.js";
+import type { XingyunAvatarAgentOptions, XingyunAvatarAgentInitParams, AgentASRStartOptions, AgentPCMChunk, AgentState, AgentASRState } from "./public-types.js";
 export * from "./public-types.js";
 export default class XingyunAvatarAgent {
     static checkCapability(audioOnlyMode?: boolean, hardwareAcceleration?: string): Promise<ICapabilityCheckReport>;
@@ -11,6 +11,8 @@ export default class XingyunAvatarAgent {
     /** @deprecated Use speak instead. */
     speakByE2E: (text: string) => Promise<void>;
     startASR: (options?: AgentASRStartOptions) => Promise<void>;
+    /** Push PCM while a startASR({ pcm }) session is active. Throws synchronously on invalid input or overflow. */
+    pushASRAudio: (chunk: AgentPCMChunk) => void;
     stopASR: () => Promise<void>;
     interrupt: () => number;
     interruptConversation: (reason?: string) => Promise<void>;

package/package-dist/agent/public-types.d.ts

@@ -30,12 +30,26 @@ export interface AgentASRResult {
     isFinal: boolean;
     raw?: unknown;
 }
+export interface AgentASRPCMInputOptions {
+    /** Input sample rate, integer in 8000-48000. */
+    sampleRate: number;
+    /** Input channels; 2 means interleaved stereo. Defaults to 1. */
+    channels?: 1 | 2;
+    /** Start-up buffer in ms (0-1000, rounded up to 100 ms). Defaults to 200. */
+    jitterBufferMs?: number;
+}
+export type AgentPCMChunk = Int16Array | Float32Array;
 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;
+    /**
+     * Developer-pushed PCM input, fed with pushASRAudio(). Mutually exclusive with inputStream.
+     * The SDK emits silence for any 100 ms slot with no pushed data.
+     */
+    pcm?: AgentASRPCMInputOptions;
 }
 export interface AgentLLMUsage {
     promptTokens: number;