Add local MeloTTS support

src/index.ts

@@ -6,6 +6,7 @@ import { Logger } from "./logger.js";
 import { printAudioDevices, spawnLoopbackCapture } from "./audio/capture.js";
 import { RealtimeSegmenter } from "./audio/realtime-segmenter.js";
 import { FasterWhisperSttService } from "./services/faster-whisper-stt.js";
+import { MeloTtsService } from "./services/melo-tts.js";
 import { OllamaLlmService } from "./services/ollama-llm.js";
 
 const mode = process.argv[2] ?? "test-stt";
@@ -15,8 +16,10 @@ async function runSttTest(enableLlm: boolean): Promise<void> {
   const logger = new Logger(config.DEBUG ? config.LOG_LEVEL : "error");
   const stt = new FasterWhisperSttService(config, logger);
   const llm = enableLlm ? new OllamaLlmService(config, logger) : null;
+  let tts = enableLlm && config.TTS_ENABLED ? new MeloTtsService(config, logger) : null;
   let capture = null as ReturnType<typeof spawnLoopbackCapture> | null;
   let shuttingDown: Promise<void> | null = null;
+  let suppressCapture = false;
   let receivedChunks = 0;
   let receivedBytes = 0;
   let maxPeak = 0;
@@ -79,6 +82,22 @@ async function runSttTest(enableLlm: boolean): Promise<void> {
     logger.info("LLM warmup finished");
     console.log("LLM 준비 완료");
   }
+  if (tts) {
+    console.log("TTS 준비중...");
+    try {
+      await tts.warmup();
+      logger.info("TTS warmup finished", {
+        image: config.TTS_IMAGE,
+        language: config.TTS_LANGUAGE,
+        speaker: config.TTS_SPEAKER,
+      });
+      console.log("TTS 준비 완료");
+    } catch (error) {
+      logger.warn("TTS warmup failed", error);
+      console.log("TTS 비활성화: bun run setup:tts 를 먼저 실행하세요.");
+      tts = null;
+    }
+  }
 
   const transcriptionQueue: Array<{ pcm16: Buffer; queuedAt: number; index: number }> = [];
   let transcribing = false;
@@ -155,6 +174,20 @@ async function runSttTest(enableLlm: boolean): Promise<void> {
           } else {
             console.log(`답변> ${reply}`);
           }
+
+          if (tts) {
+            suppressCapture = true;
+            segmenter.reset();
+            try {
+              await tts.speak(reply);
+            } catch (error) {
+              logger.warn("TTS playback failed", error);
+            } finally {
+              suppressCapture = false;
+              sawSpeechStart = false;
+              maxPeak = 0;
+            }
+          }
         }
       }
     } catch (error) {
@@ -227,6 +260,9 @@ async function runSttTest(enableLlm: boolean): Promise<void> {
     receivedChunks += 1;
     receivedBytes += chunk.length;
     lastChunkAt = Date.now();
+    if (suppressCapture) {
+      return;
+    }
     segmenter.pushChunk(chunk);
   });
   capture.stderr.on("data", (chunk: Buffer) => {
@@ -330,6 +366,19 @@ async function runLlmCli(): Promise<void> {
   });
 }
 
+async function runTtsTest(): Promise<void> {
+  const text = process.argv.slice(3).join(" ").trim() || "안녕하세요. 로컬 티티에스 테스트입니다.";
+  const config = loadConfig();
+  const logger = new Logger(config.DEBUG ? config.LOG_LEVEL : "error");
+  const tts = new MeloTtsService(config, logger);
+
+  console.log("TTS 준비중...");
+  await tts.warmup();
+  console.log("TTS 준비 완료");
+  console.log(`재생 문장: ${text}`);
+  await tts.speak(text);
+}
+
 async function main(): Promise<void> {
   switch (mode) {
     case "devices":
@@ -344,8 +393,11 @@ async function main(): Promise<void> {
     case "test-llm":
       await runLlmCli();
       return;
+    case "test-tts":
+      await runTtsTest();
+      return;
     default:
-      throw new Error(`알 수 없는 실행 모드입니다: ${mode}. 사용 가능: test-stt, test-sttllm, test-llm, devices`);
+      throw new Error(`알 수 없는 실행 모드입니다: ${mode}. 사용 가능: test-stt, test-sttllm, test-llm, test-tts, devices`);
   }
 }