@dtelecom/agents-js 0.1.5 → 0.1.8

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-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';
+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-f6SAlHpW.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-f6SAlHpW.mjs';
 
 declare class VoiceAgent extends EventEmitter {
     private readonly config;
@@ -103,6 +103,8 @@ declare class Pipeline extends EventEmitter {
     private readonly nameVariants;
     private readonly beforeRespond?;
     private readonly memory?;
+    /** Strip provider-specific markup (e.g. SSML lang tags) for display. */
+    private cleanText;
     /** Active STT streams, keyed by participant identity */
     private sttStreams;
     private _processing;

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-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';
+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-f6SAlHpW.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-f6SAlHpW.js';
 
 declare class VoiceAgent extends EventEmitter {
     private readonly config;
@@ -103,6 +103,8 @@ declare class Pipeline extends EventEmitter {
     private readonly nameVariants;
     private readonly beforeRespond?;
     private readonly memory?;
+    /** Strip provider-specific markup (e.g. SSML lang tags) for display. */
+    private cleanText;
     /** Active STT streams, keyed by participant identity */
     private sttStreams;
     private _processing;

package/dist/index.js

@@ -1156,6 +1156,10 @@ var Pipeline = class extends import_events.EventEmitter {
   nameVariants;
   beforeRespond;
   memory;
+  /** Strip provider-specific markup (e.g. SSML lang tags) for display. */
+  cleanText(text) {
+    return this.tts?.cleanText ? this.tts.cleanText(text) : text;
+  }
   /** Active STT streams, keyed by participant identity */
   sttStreams = /* @__PURE__ */ new Map();
   _processing = false;
@@ -1404,12 +1408,16 @@ var Pipeline = class extends import_events.EventEmitter {
                 log7.debug(`Skipping non-word sentence: "${sentence}"`);
                 continue;
               }
-              await this.synthesizeAndPlay(sentence, signal, (t) => {
+              let processed = sentence;
+              if (this.tts?.preprocessText) {
+                processed = await this.tts.preprocessText(sentence, signal);
+              }
+              await this.synthesizeAndPlay(processed, signal, (t) => {
                 if (!tFirstAudioPlayed) {
                   tFirstAudioPlayed = t;
                   this.setAgentState("speaking");
                 }
-                this.emit("sentence", sentence);
+                this.emit("sentence", this.cleanText(processed));
               });
               continue;
             }
@@ -1437,7 +1445,7 @@ var Pipeline = class extends import_events.EventEmitter {
       if (fullResponse.trim()) {
         this.context.addAgentTurn(fullResponse.trim());
         this.memory?.storeTurn("assistant", fullResponse.trim(), true);
-        this.emit("response", fullResponse.trim());
+        this.emit("response", this.cleanText(fullResponse.trim()));
       }
       await sleep2(AUDIO_DRAIN_MS);
       this.setAgentState("idle");
@@ -1476,15 +1484,19 @@ var Pipeline = class extends import_events.EventEmitter {
       const signal = this.bargeIn.startCycle();
       this.audioOutput.beginResponse();
       this.setAgentState("thinking");
-      await this.synthesizeAndPlay(text, signal, () => {
+      let processed = text;
+      if (this.tts?.preprocessText) {
+        processed = await this.tts.preprocessText(text, signal);
+      }
+      await this.synthesizeAndPlay(processed, signal, () => {
         this.setAgentState("speaking");
-        this.emit("sentence", text);
+        this.emit("sentence", this.cleanText(processed));
       });
       if (!signal.aborted) {
         await this.audioOutput.writeSilence(40);
         this.context.addAgentTurn(text);
         this.memory?.storeTurn("assistant", text, true);
-        this.emit("response", text);
+        this.emit("response", this.cleanText(processed));
       }
       await sleep2(AUDIO_DRAIN_MS);
       this.setAgentState("idle");