@dtelecom/agents-js 0.3.0 → 0.3.1

package/dist/index.d.mts

@@ -1,8 +1,8 @@
 import * as _dtelecom_server_sdk_node from '@dtelecom/server-sdk-node';
 import { Room, AudioSource, RemoteAudioTrack, AudioFrame } from '@dtelecom/server-sdk-node';
 import { EventEmitter } from 'events';
-import { A as AgentConfig, a as AgentStartOptions, M as Message, L as LLMPlugin, P as PipelineOptions, b as AgentState, S as STTStream, T as TranscriptionResult } from './types-Di_jxIgs.mjs';
-export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, e as LLMChatOptions, f as LLMChunk, g as MemoryConfig, h as PipelineEvents, R as RespondMode, i as STTPlugin, j as STTStreamOptions, k as TTSPlugin, l as ToolCallResult, m as ToolDefinition } from './types-Di_jxIgs.mjs';
+import { A as AgentConfig, a as AgentStartOptions, M as Message, L as LLMPlugin, P as PipelineOptions, b as AgentState, S as STTStream, T as TranscriptionResult } from './types-BJylZd8Q.mjs';
+export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, e as LLMChatOptions, f as LLMChunk, g as MemoryConfig, h as PipelineEvents, R as RespondMode, i as STTPlugin, j as STTStreamOptions, k as TTSPlugin, l as ToolCallResult, m as ToolDefinition } from './types-BJylZd8Q.mjs';
 
 declare class VoiceAgent extends EventEmitter {
     private readonly config;
@@ -125,6 +125,7 @@ declare class Pipeline extends EventEmitter {
     private readonly _warmupPromise;
     private readonly _ttsWarmupPromise;
     private readonly _llmWarmupPromise;
+    private readonly _audioReadyPromise;
     get processing(): boolean;
     get running(): boolean;
     get agentState(): AgentState;

package/dist/index.d.ts

@@ -1,8 +1,8 @@
 import * as _dtelecom_server_sdk_node from '@dtelecom/server-sdk-node';
 import { Room, AudioSource, RemoteAudioTrack, AudioFrame } from '@dtelecom/server-sdk-node';
 import { EventEmitter } from 'events';
-import { A as AgentConfig, a as AgentStartOptions, M as Message, L as LLMPlugin, P as PipelineOptions, b as AgentState, S as STTStream, T as TranscriptionResult } from './types-Di_jxIgs.js';
-export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, e as LLMChatOptions, f as LLMChunk, g as MemoryConfig, h as PipelineEvents, R as RespondMode, i as STTPlugin, j as STTStreamOptions, k as TTSPlugin, l as ToolCallResult, m as ToolDefinition } from './types-Di_jxIgs.js';
+import { A as AgentConfig, a as AgentStartOptions, M as Message, L as LLMPlugin, P as PipelineOptions, b as AgentState, S as STTStream, T as TranscriptionResult } from './types-BJylZd8Q.js';
+export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, e as LLMChatOptions, f as LLMChunk, g as MemoryConfig, h as PipelineEvents, R as RespondMode, i as STTPlugin, j as STTStreamOptions, k as TTSPlugin, l as ToolCallResult, m as ToolDefinition } from './types-BJylZd8Q.js';
 
 declare class VoiceAgent extends EventEmitter {
     private readonly config;
@@ -125,6 +125,7 @@ declare class Pipeline extends EventEmitter {
     private readonly _warmupPromise;
     private readonly _ttsWarmupPromise;
     private readonly _llmWarmupPromise;
+    private readonly _audioReadyPromise;
     get processing(): boolean;
     get running(): boolean;
     get agentState(): AgentState;

package/dist/index.js

@@ -682,11 +682,21 @@ var AudioOutput = class {
   _responding = false;
   _stopped = false;
   silenceInterval = null;
+  /** Resolves when the RTP transport is ready and initial silence has been sent. */
+  whenReady;
+  _resolveReady;
   /** When set, raw PCM from TTS is saved to this directory as WAV files for debugging. */
   dumpDir = null;
   dumpCounter = 0;
   constructor(source) {
     this.source = source;
+    if (source.ready) {
+      this.whenReady = Promise.resolve();
+    } else {
+      this.whenReady = new Promise((resolve) => {
+        this._resolveReady = resolve;
+      });
+    }
   }
   get playing() {
     return this._playing;
@@ -714,8 +724,11 @@ var AudioOutput = class {
   startSilence() {
     if (this.silenceInterval) return;
     const startKeepalive = () => {
-      log3.debug("Transport ready \u2014 sending initial silence + starting 3s keepalive");
-      this.sendSilenceFrame();
+      log3.debug("Transport ready \u2014 sending initial silence burst + starting 3s keepalive");
+      for (let i = 0; i < 15; i++) {
+        this.sendSilenceFrame();
+      }
+      this._resolveReady?.();
       this.silenceInterval = setInterval(() => {
         if (!this._playing && !this._responding && !this._stopped) {
           this.sendSilenceFrame();
@@ -1254,12 +1267,14 @@ var Pipeline = class extends import_events.EventEmitter {
     this._llmWarmupPromise = this.llm.warmup ? this.llm.warmup(options.instructions).catch((err) => {
       log7.warn("LLM warmup failed (non-fatal):", err);
     }) : Promise.resolve();
-    this._warmupPromise = Promise.all([this._ttsWarmupPromise, this._llmWarmupPromise]).then(() => {
+    this._audioReadyPromise = this.audioOutput.whenReady;
+    this._warmupPromise = Promise.all([this._ttsWarmupPromise, this._llmWarmupPromise, this._audioReadyPromise]).then(() => {
     });
   }
   _warmupPromise;
   _ttsWarmupPromise;
   _llmWarmupPromise;
+  _audioReadyPromise;
   get processing() {
     return this._processing;
   }
@@ -1417,6 +1432,7 @@ var Pipeline = class extends import_events.EventEmitter {
         wake();
       };
       const MAX_LLM_RETRIES = 2;
+      let toolCallEmitted = false;
       const producer = async () => {
         const defaultLang = this.tts?.defaultLanguage;
         for (let attempt = 0; attempt <= MAX_LLM_RETRIES; attempt++) {
@@ -1466,6 +1482,7 @@ var Pipeline = class extends import_events.EventEmitter {
                 }
               } else if (chunk.type === "tool_call" && chunk.toolCall) {
                 log7.info(`Tool call: ${chunk.toolCall.name}(${chunk.toolCall.arguments})`);
+                toolCallEmitted = true;
                 this.emit("toolCall", chunk.toolCall);
               }
             }
@@ -1478,7 +1495,7 @@ var Pipeline = class extends import_events.EventEmitter {
             if (remaining) {
               pushSentence(remaining);
             }
-            if (fullResponse.trim()) {
+            if (fullResponse.trim() || toolCallEmitted) {
               break;
             }
             log7.warn(`LLM produced no output (attempt ${attempt + 1}/${MAX_LLM_RETRIES + 1})`);
@@ -1590,7 +1607,7 @@ var Pipeline = class extends import_events.EventEmitter {
       return;
     }
     this._processing = true;
-    await this._ttsWarmupPromise;
+    await Promise.all([this._ttsWarmupPromise, this._audioReadyPromise]);
     log7.info(`say(): "${text.slice(0, 60)}"`);
     try {
       const signal = this.bargeIn.startCycle();