Warn when Windows local input is silent

src/audio/local-voice-session.ts

@@ -33,6 +33,7 @@ export class LocalVoiceSession {
   private readonly memory: ConversationMemory;
   private readonly queue: SpeechJob[] = [];
   private readonly pendingSamples: number[] = [];
+  private readonly silenceThreshold = 900;
 
   private vad: RealTimeVAD | null = null;
   private recorder: ChildProcessByStdio<null, Readable, Readable> | null = null;
@@ -42,6 +43,12 @@ export class LocalVoiceSession {
   private processing = Promise.resolve();
   private draining = false;
   private destroyed = false;
+  private inputWatchdog: NodeJS.Timeout | null = null;
+  private recorderStartedAt = 0;
+  private lastPcmChunkAt = 0;
+  private lastNonSilentAudioAt = 0;
+  private warnedNoPcm = false;
+  private warnedSilent = false;
 
   constructor(private readonly options: LocalVoiceSessionOptions) {
     this.memory = new ConversationMemory(options.config.MAX_CONVERSATION_TURNS);
@@ -69,6 +76,11 @@ export class LocalVoiceSession {
     });
 
     this.recorder = this.spawnRecorder();
+    this.recorderStartedAt = Date.now();
+    this.lastPcmChunkAt = 0;
+    this.lastNonSilentAudioAt = 0;
+    this.warnedNoPcm = false;
+    this.warnedSilent = false;
     this.recorder.stdout.on("data", (chunk: Buffer) => {
       this.pushPcm16Chunk(chunk);
     });
@@ -83,12 +95,21 @@ export class LocalVoiceSession {
         this.options.logger.warn("pw-record exited unexpectedly", { code, signal });
       }
     });
+
+    this.inputWatchdog = setInterval(() => {
+      this.reportInputHealth();
+    }, 3_000);
   }
 
   async destroy(): Promise<void> {
     this.destroyed = true;
     this.interruptPlayback("local-shutdown");
 
+    if (this.inputWatchdog) {
+      clearInterval(this.inputWatchdog);
+      this.inputWatchdog = null;
+    }
+
     if (this.recorder && !this.recorder.killed) {
       this.recorder.kill("SIGTERM");
       await once(this.recorder, "exit").catch(() => null);
@@ -195,8 +216,20 @@ export class LocalVoiceSession {
       return;
     }
 
+    this.lastPcmChunkAt = Date.now();
+    let peak = 0;
+
     for (let offset = 0; offset + 1 < chunk.length; offset += 2) {
-      this.pendingSamples.push(chunk.readInt16LE(offset));
+      const sample = chunk.readInt16LE(offset);
+      const abs = Math.abs(sample);
+      if (abs > peak) {
+        peak = abs;
+      }
+      this.pendingSamples.push(sample);
+    }
+
+    if (peak >= this.silenceThreshold) {
+      this.lastNonSilentAudioAt = Date.now();
     }
 
     while (true) {
@@ -216,6 +249,7 @@ export class LocalVoiceSession {
 
   private async handleSpeechEnd(audio: Float32Array): Promise<void> {
     if (audio.length < 16000 * 0.25) {
+      this.options.logger.debug("Ignored short local speech segment", { samples: audio.length });
       return;
     }
 
@@ -234,6 +268,7 @@ export class LocalVoiceSession {
     }
 
     if (!transcript || transcript.trim().length === 0) {
+      this.options.logger.info("Local STT returned empty transcript");
       return;
     }
 
@@ -463,6 +498,39 @@ export class LocalVoiceSession {
     return requireFfmpegPath();
   }
 
+  private reportInputHealth(): void {
+    if (this.destroyed) {
+      return;
+    }
+
+    const now = Date.now();
+
+    if (!this.warnedNoPcm && this.lastPcmChunkAt === 0 && now - this.recorderStartedAt >= 6_000) {
+      this.warnedNoPcm = true;
+      this.options.logger.warn(
+        [
+          "입력 장치에서 PCM 데이터가 들어오지 않습니다.",
+          `현재 source: ${this.options.config.LOCAL_AUDIO_SOURCE ?? "default"}`,
+          "Windows에서는 마이크 입력이 아니라 SPDIF/ADAT 같은 디지털 입력을 고르면 반응이 없습니다.",
+          "`bun run devices`로 실제 마이크 이름을 다시 고르세요.",
+        ].join("\n"),
+      );
+      return;
+    }
+
+    if (!this.warnedSilent && this.lastPcmChunkAt > 0 && this.lastNonSilentAudioAt === 0 && now - this.recorderStartedAt >= 6_000) {
+      this.warnedSilent = true;
+      this.options.logger.warn(
+        [
+          "입력 장치에서는 데이터가 오지만 말소리 수준으로 올라오지 않습니다.",
+          `현재 source: ${this.options.config.LOCAL_AUDIO_SOURCE ?? "default"}`,
+          "잘못된 입력 채널이거나, 마이크가 그 장치로 라우팅되지 않은 상태일 가능성이 큽니다.",
+          "RME Babyface Pro라면 SPDIF/ADAT 대신 아날로그 마이크 입력 채널 이름을 선택해야 합니다.",
+        ].join("\n"),
+      );
+    }
+  }
+
   private describeSink(): string {
     if (process.platform === "win32") {
       return this.options.config.LOCAL_AUDIO_SINK ?? "system-default";