@dtelecom/agents-js 0.1.2 → 0.1.5

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-EvtHMokR.mjs';
-export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, e as LLMChunk, f as MemoryConfig, g as PipelineEvents, R as RespondMode, h as STTPlugin, i as STTStreamOptions, j as TTSPlugin } from './types-EvtHMokR.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-DWdkYmW8.mjs';
+export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, e as LLMChunk, f as MemoryConfig, g as PipelineEvents, R as RespondMode, h as STTPlugin, i as STTStreamOptions, j as TTSPlugin } from './types-DWdkYmW8.mjs';
 
 declare class VoiceAgent extends EventEmitter {
     private readonly config;
@@ -111,6 +111,9 @@ declare class Pipeline extends EventEmitter {
     /** Queued turn while current one is still processing */
     private pendingTurn;
     constructor(options: PipelineOptions);
+    /** One-shot warmup — safe to call from constructor, resolves when both LLM and TTS are ready. */
+    private _warmupPromise;
+    private warmup;
     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-EvtHMokR.js';
-export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, e as LLMChunk, f as MemoryConfig, g as PipelineEvents, R as RespondMode, h as STTPlugin, i as STTStreamOptions, j as TTSPlugin } from './types-EvtHMokR.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-DWdkYmW8.js';
+export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, e as LLMChunk, f as MemoryConfig, g as PipelineEvents, R as RespondMode, h as STTPlugin, i as STTStreamOptions, j as TTSPlugin } from './types-DWdkYmW8.js';
 
 declare class VoiceAgent extends EventEmitter {
     private readonly config;
@@ -111,6 +111,9 @@ declare class Pipeline extends EventEmitter {
     /** Queued turn while current one is still processing */
     private pendingTurn;
     constructor(options: PipelineOptions);
+    /** One-shot warmup — safe to call from constructor, resolves when both LLM and TTS are ready. */
+    private _warmupPromise;
+    private warmup;
     get processing(): boolean;
     get running(): boolean;
     get agentState(): AgentState;

package/dist/index.js

@@ -698,17 +698,33 @@ var AudioOutput = class {
    * Start sparse silence keepalive to prevent the SFU from dropping the track.
    * With Opus DTX enabled, the encoder handles silence natively — we only need
    * an occasional packet to keep the SSRC alive.
+   *
+   * Waits for the RTP transport to be ready before sending — no frames are
+   * wasted before DTLS is connected.
    */
   startSilence() {
     if (this.silenceInterval) return;
-    log3.debug("Starting silence keepalive (sparse, 3s interval)");
-    this.silenceInterval = setInterval(() => {
-      if (!this._playing && !this._responding && !this._stopped) {
-        const f = new import_server_sdk_node2.AudioFrame(SILENCE, SAMPLE_RATE, CHANNELS, SAMPLES_PER_FRAME);
-        this.source.captureFrame(f).catch(() => {
-        });
-      }
-    }, 3e3);
+    const startKeepalive = () => {
+      log3.debug("Transport ready \u2014 sending initial silence + starting 3s keepalive");
+      this.sendSilenceFrame();
+      this.silenceInterval = setInterval(() => {
+        if (!this._playing && !this._responding && !this._stopped) {
+          this.sendSilenceFrame();
+        }
+      }, 3e3);
+    };
+    if (this.source.ready) {
+      startKeepalive();
+    } else {
+      log3.debug("Waiting for transport before starting silence keepalive...");
+      this.source.onReady = () => startKeepalive();
+    }
+  }
+  sendSilenceFrame() {
+    const frame = new import_server_sdk_node2.AudioFrame(SILENCE, SAMPLE_RATE, CHANNELS, SAMPLES_PER_FRAME);
+    this.source.captureFrame(frame).catch((err) => {
+      log3.warn("Failed to send silence frame:", err);
+    });
   }
   /**
    * Write a PCM16 buffer to the audio output.
@@ -1173,16 +1189,27 @@ var Pipeline = class extends import_events.EventEmitter {
       this.splitter.reset();
       this.setAgentState("idle");
     };
+    this._warmupPromise = this.warmup(options.instructions);
+  }
+  /** One-shot warmup — safe to call from constructor, resolves when both LLM and TTS are ready. */
+  _warmupPromise;
+  async warmup(instructions) {
+    const tasks = [];
     if (this.llm.warmup) {
-      this.llm.warmup(options.instructions).catch((err) => {
-        log7.warn("LLM warmup failed:", err);
-      });
+      tasks.push(
+        this.llm.warmup(instructions).catch((err) => {
+          log7.warn("LLM warmup failed:", err);
+        })
+      );
     }
     if (this.tts?.warmup) {
-      this.tts.warmup().catch((err) => {
-        log7.warn("TTS warmup failed:", err);
-      });
+      tasks.push(
+        this.tts.warmup().catch((err) => {
+          log7.warn("TTS warmup failed:", err);
+        })
+      );
     }
+    await Promise.all(tasks);
   }
   get processing() {
     return this._processing;
@@ -1289,6 +1316,7 @@ var Pipeline = class extends import_events.EventEmitter {
       return;
     }
     this._processing = true;
+    await this._warmupPromise;
     const tSpeechEnd = this.lastFinalAt;
     const sttDuration = this.lastSttDuration;
     let tLlmFirstToken = 0;
@@ -1442,6 +1470,7 @@ var Pipeline = class extends import_events.EventEmitter {
       return;
     }
     this._processing = true;
+    await this._warmupPromise;
     log7.info(`say(): "${text.slice(0, 60)}"`);
     try {
       const signal = this.bargeIn.startCycle();