@pompeii-labs/audio 0.0.3 → 0.0.4

package/dist/voice.d.mts

@@ -15,9 +15,9 @@ declare abstract class MagmaFlowSpeechToText {
 }
 
 declare abstract class MagmaFlowTextToSpeech {
-    abstract input(text: string): void;
+    abstract input(text: string, requestId: string): void;
     abstract kill(): void;
-    onOutput(audio: Buffer | null): void;
+    onOutput(audio: Buffer | null, requestId: string): void;
     constructor();
 }
 
@@ -44,13 +44,15 @@ declare class MagmaFlow {
     private textBuffer;
     private textQueue;
     private generatingAudio;
+    private currentRequestId;
     private audioBuffer;
     private config;
     constructor(args: MagmaFlowArgs);
     inputAudio(audio: Buffer): void;
-    inputText(text: string): void;
+    inputText(text: string | null): void;
     private generateAudio;
     private sendAudio;
+    interruptTTS(): void;
     kill(): void;
 }
 
@@ -88,8 +90,7 @@ declare class DeepgramTTS extends MagmaFlowTextToSpeech {
     private client;
     constructor(args: DeepgramTTSArgs);
     setup(): Promise<void>;
-    input(text: string | null): void;
-    private output;
+    input(text: string | null, requestId: string): void;
     kill(): void;
     reset(): void;
 }
@@ -115,8 +116,7 @@ declare class ElevenLabsTTS extends MagmaFlowTextToSpeech {
     private config;
     constructor(args: ElevenLabsTTSArgs);
     setup(): Promise<void>;
-    input(text: string | null): void;
-    private output;
+    input(text: string | null, requestId: string): void;
     kill(): void;
     reset(): void;
 }
@@ -128,8 +128,7 @@ declare class HumeTTS extends MagmaFlowTextToSpeech {
     private client;
     constructor(args: HumeTTSArgs);
     setup(): Promise<void>;
-    input(text: string | null): void;
-    private output;
+    input(text: string | null, requestId: string): void;
     kill(): void;
     reset(): void;
 }
@@ -141,8 +140,7 @@ declare class WhisperTTS extends MagmaFlowTextToSpeech {
     private client;
     constructor(args: WhisperTTSArgs);
     setup(): Promise<void>;
-    input(text: string | null): void;
-    private output;
+    input(text: string | null, requestId: string): void;
     kill(): void;
     reset(): void;
 }

package/dist/voice.d.ts

@@ -15,9 +15,9 @@ declare abstract class MagmaFlowSpeechToText {
 }
 
 declare abstract class MagmaFlowTextToSpeech {
-    abstract input(text: string): void;
+    abstract input(text: string, requestId: string): void;
     abstract kill(): void;
-    onOutput(audio: Buffer | null): void;
+    onOutput(audio: Buffer | null, requestId: string): void;
     constructor();
 }
 
@@ -44,13 +44,15 @@ declare class MagmaFlow {
     private textBuffer;
     private textQueue;
     private generatingAudio;
+    private currentRequestId;
     private audioBuffer;
     private config;
     constructor(args: MagmaFlowArgs);
     inputAudio(audio: Buffer): void;
-    inputText(text: string): void;
+    inputText(text: string | null): void;
     private generateAudio;
     private sendAudio;
+    interruptTTS(): void;
     kill(): void;
 }
 
@@ -88,8 +90,7 @@ declare class DeepgramTTS extends MagmaFlowTextToSpeech {
     private client;
     constructor(args: DeepgramTTSArgs);
     setup(): Promise<void>;
-    input(text: string | null): void;
-    private output;
+    input(text: string | null, requestId: string): void;
     kill(): void;
     reset(): void;
 }
@@ -115,8 +116,7 @@ declare class ElevenLabsTTS extends MagmaFlowTextToSpeech {
     private config;
     constructor(args: ElevenLabsTTSArgs);
     setup(): Promise<void>;
-    input(text: string | null): void;
-    private output;
+    input(text: string | null, requestId: string): void;
     kill(): void;
     reset(): void;
 }
@@ -128,8 +128,7 @@ declare class HumeTTS extends MagmaFlowTextToSpeech {
     private client;
     constructor(args: HumeTTSArgs);
     setup(): Promise<void>;
-    input(text: string | null): void;
-    private output;
+    input(text: string | null, requestId: string): void;
     kill(): void;
     reset(): void;
 }
@@ -141,8 +140,7 @@ declare class WhisperTTS extends MagmaFlowTextToSpeech {
     private client;
     constructor(args: WhisperTTSArgs);
     setup(): Promise<void>;
-    input(text: string | null): void;
-    private output;
+    input(text: string | null, requestId: string): void;
     kill(): void;
     reset(): void;
 }

package/dist/voice.js

@@ -391,6 +391,7 @@ var MagmaFlow = class {
     this.textBuffer = "";
     this.textQueue = [];
     this.generatingAudio = false;
+    this.currentRequestId = null;
     this.audioBuffer = [];
     this.config = {
       pauseDurationMs: 500,
@@ -402,7 +403,11 @@ var MagmaFlow = class {
     this.outputFormat = args.outputFormat;
     this.onAudioOutput = args.onAudioOutput;
     this.config = { ...this.config, ...args.config };
-    this.tts.onOutput = (audio) => {
+    this.tts.onOutput = (audio, requestId) => {
+      if (this.currentRequestId !== requestId) {
+        console.log("[MagmaFlow] Skipping output for cancelled request");
+        return;
+      }
       if (!audio) {
         const lastChunk = this.audioBuffer[this.audioBuffer.length - 1];
         if (lastChunk) {
@@ -453,7 +458,10 @@ var MagmaFlow = class {
     const chunk = this.textQueue.shift();
     if (!chunk) return;
     this.generatingAudio = true;
-    this.tts.input(chunk);
+    if (!this.currentRequestId) {
+      this.currentRequestId = crypto.randomUUID();
+    }
+    this.tts.input(chunk, this.currentRequestId);
   }
   sendAudio() {
     if (this.audioBuffer.length === 0) return;
@@ -471,6 +479,13 @@ var MagmaFlow = class {
       console.error("Audio output callback error:", error);
     }
   }
+  interruptTTS() {
+    this.textQueue = [];
+    this.textBuffer = "";
+    this.audioBuffer = [];
+    this.generatingAudio = false;
+    this.currentRequestId = null;
+  }
   kill() {
     this.stt.kill();
     this.tts.kill();
@@ -587,7 +602,7 @@ var DeepgramSTT = class extends MagmaFlowSpeechToText {
 
 // src/voice/textToSpeech/base.ts
 var MagmaFlowTextToSpeech = class {
-  onOutput(audio) {
+  onOutput(audio, requestId) {
     console.log("[Default TTS] Output:", audio);
   }
   constructor() {
@@ -600,7 +615,7 @@ var DeepgramTTS = class extends MagmaFlowTextToSpeech {
   }
   async setup() {
   }
-  input(text) {
+  input(text, requestId) {
     if (!text) {
       return;
     }
@@ -620,15 +635,12 @@ var DeepgramTTS = class extends MagmaFlowTextToSpeech {
         return;
       }
       for await (const chunk of stream) {
-        this.output(Buffer.from(chunk));
+        this.onOutput(Buffer.from(chunk), requestId);
       }
-      this.output(null);
+      this.onOutput(null, requestId);
       console.log("[Deepgram] Finished:", text);
     });
   }
-  output(audio) {
-    this.onOutput(audio);
-  }
   kill() {
   }
   reset() {
@@ -654,7 +666,7 @@ var ElevenLabsTTS = class extends MagmaFlowTextToSpeech {
   }
   async setup() {
   }
-  input(text) {
+  input(text, requestId) {
     if (!text) {
       return;
     }
@@ -665,15 +677,12 @@ var ElevenLabsTTS = class extends MagmaFlowTextToSpeech {
       ...this.config
     }).then(async (stream) => {
       for await (const chunk of stream) {
-        this.output(chunk);
+        this.onOutput(chunk, requestId);
       }
-      this.output(null);
+      this.onOutput(null, requestId);
       console.log("[ElevenLabs] Finished:", text);
     });
   }
-  output(audio) {
-    this.onOutput(audio);
-  }
   kill() {
   }
   reset() {
@@ -686,7 +695,7 @@ var HumeTTS = class extends MagmaFlowTextToSpeech {
   }
   async setup() {
   }
-  input(text) {
+  input(text, requestId) {
     if (!text) {
       return;
     }
@@ -702,15 +711,12 @@ var HumeTTS = class extends MagmaFlowTextToSpeech {
       instantMode: true
     }).then(async (stream) => {
       for await (const chunk of stream) {
-        this.output(Buffer.from(chunk.audio, "base64"));
+        this.onOutput(Buffer.from(chunk.audio, "base64"), requestId);
       }
-      this.output(null);
+      this.onOutput(null, requestId);
       console.log("[Hume] Finished:", text);
     });
   }
-  output(audio) {
-    this.onOutput(audio);
-  }
   kill() {
   }
   reset() {
@@ -723,7 +729,7 @@ var WhisperTTS = class extends MagmaFlowTextToSpeech {
   }
   async setup() {
   }
-  input(text) {
+  input(text, requestId) {
     if (!text) {
       return;
     }
@@ -739,14 +745,11 @@ var WhisperTTS = class extends MagmaFlowTextToSpeech {
         24e3,
         48e3
       );
-      this.output(int16ArrayToBuffer(resampledPCM));
-      this.output(null);
+      this.onOutput(int16ArrayToBuffer(resampledPCM), requestId);
+      this.onOutput(null, requestId);
       console.log("[Whisper] Finished:", text);
     });
   }
-  output(audio) {
-    this.onOutput(audio);
-  }
   kill() {
   }
   reset() {

package/dist/voice.mjs

@@ -385,6 +385,7 @@ var MagmaFlow = class {
     this.textBuffer = "";
     this.textQueue = [];
     this.generatingAudio = false;
+    this.currentRequestId = null;
     this.audioBuffer = [];
     this.config = {
       pauseDurationMs: 500,
@@ -396,7 +397,11 @@ var MagmaFlow = class {
     this.outputFormat = args.outputFormat;
     this.onAudioOutput = args.onAudioOutput;
     this.config = { ...this.config, ...args.config };
-    this.tts.onOutput = (audio) => {
+    this.tts.onOutput = (audio, requestId) => {
+      if (this.currentRequestId !== requestId) {
+        console.log("[MagmaFlow] Skipping output for cancelled request");
+        return;
+      }
       if (!audio) {
         const lastChunk = this.audioBuffer[this.audioBuffer.length - 1];
         if (lastChunk) {
@@ -447,7 +452,10 @@ var MagmaFlow = class {
     const chunk = this.textQueue.shift();
     if (!chunk) return;
     this.generatingAudio = true;
-    this.tts.input(chunk);
+    if (!this.currentRequestId) {
+      this.currentRequestId = crypto.randomUUID();
+    }
+    this.tts.input(chunk, this.currentRequestId);
   }
   sendAudio() {
     if (this.audioBuffer.length === 0) return;
@@ -465,6 +473,13 @@ var MagmaFlow = class {
       console.error("Audio output callback error:", error);
     }
   }
+  interruptTTS() {
+    this.textQueue = [];
+    this.textBuffer = "";
+    this.audioBuffer = [];
+    this.generatingAudio = false;
+    this.currentRequestId = null;
+  }
   kill() {
     this.stt.kill();
     this.tts.kill();
@@ -581,7 +596,7 @@ var DeepgramSTT = class extends MagmaFlowSpeechToText {
 
 // src/voice/textToSpeech/base.ts
 var MagmaFlowTextToSpeech = class {
-  onOutput(audio) {
+  onOutput(audio, requestId) {
     console.log("[Default TTS] Output:", audio);
   }
   constructor() {
@@ -594,7 +609,7 @@ var DeepgramTTS = class extends MagmaFlowTextToSpeech {
   }
   async setup() {
   }
-  input(text) {
+  input(text, requestId) {
     if (!text) {
       return;
     }
@@ -614,15 +629,12 @@ var DeepgramTTS = class extends MagmaFlowTextToSpeech {
         return;
       }
       for await (const chunk of stream) {
-        this.output(Buffer.from(chunk));
+        this.onOutput(Buffer.from(chunk), requestId);
       }
-      this.output(null);
+      this.onOutput(null, requestId);
       console.log("[Deepgram] Finished:", text);
     });
   }
-  output(audio) {
-    this.onOutput(audio);
-  }
   kill() {
   }
   reset() {
@@ -648,7 +660,7 @@ var ElevenLabsTTS = class extends MagmaFlowTextToSpeech {
   }
   async setup() {
   }
-  input(text) {
+  input(text, requestId) {
     if (!text) {
       return;
     }
@@ -659,15 +671,12 @@ var ElevenLabsTTS = class extends MagmaFlowTextToSpeech {
       ...this.config
     }).then(async (stream) => {
       for await (const chunk of stream) {
-        this.output(chunk);
+        this.onOutput(chunk, requestId);
       }
-      this.output(null);
+      this.onOutput(null, requestId);
       console.log("[ElevenLabs] Finished:", text);
     });
   }
-  output(audio) {
-    this.onOutput(audio);
-  }
   kill() {
   }
   reset() {
@@ -680,7 +689,7 @@ var HumeTTS = class extends MagmaFlowTextToSpeech {
   }
   async setup() {
   }
-  input(text) {
+  input(text, requestId) {
     if (!text) {
       return;
     }
@@ -696,15 +705,12 @@ var HumeTTS = class extends MagmaFlowTextToSpeech {
       instantMode: true
     }).then(async (stream) => {
       for await (const chunk of stream) {
-        this.output(Buffer.from(chunk.audio, "base64"));
+        this.onOutput(Buffer.from(chunk.audio, "base64"), requestId);
       }
-      this.output(null);
+      this.onOutput(null, requestId);
       console.log("[Hume] Finished:", text);
     });
   }
-  output(audio) {
-    this.onOutput(audio);
-  }
   kill() {
   }
   reset() {
@@ -717,7 +723,7 @@ var WhisperTTS = class extends MagmaFlowTextToSpeech {
   }
   async setup() {
   }
-  input(text) {
+  input(text, requestId) {
     if (!text) {
       return;
     }
@@ -733,14 +739,11 @@ var WhisperTTS = class extends MagmaFlowTextToSpeech {
         24e3,
         48e3
       );
-      this.output(int16ArrayToBuffer(resampledPCM));
-      this.output(null);
+      this.onOutput(int16ArrayToBuffer(resampledPCM), requestId);
+      this.onOutput(null, requestId);
       console.log("[Whisper] Finished:", text);
     });
   }
-  output(audio) {
-    this.onOutput(audio);
-  }
   kill() {
   }
   reset() {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pompeii-labs/audio",
-  "version": "0.0.3",
+  "version": "0.0.4",
   "description": "The Audio SDK from Pompeii Labs",
   "keywords": [
     "Pompeii",