pi-voice 0.1.0 → 0.2.0

package/out/preload/index.cjs

@@ -7,8 +7,6 @@ const IPC = {
   PLAY_AUDIO_STREAM_START: "play-audio-stream-start",
   PLAY_AUDIO_STREAM_CHUNK: "play-audio-stream-chunk",
   PLAY_AUDIO_STREAM_END: "play-audio-stream-end",
-  STATE_CHANGED: "state-changed",
-  STATUS_MESSAGE: "status-message",
   // renderer -> main
   RECORDING_DATA: "recording-data",
   RECORDING_ERROR: "recording-error",
@@ -16,7 +14,7 @@ const IPC = {
 };
 const api = {
   onStartRecording: (callback) => {
-    electron.ipcRenderer.on(IPC.START_RECORDING, () => callback());
+    electron.ipcRenderer.on(IPC.START_RECORDING, (_event, format) => callback(format ?? "webm"));
   },
   onStopRecording: (callback) => {
     electron.ipcRenderer.on(IPC.STOP_RECORDING, () => callback());
@@ -36,12 +34,6 @@ const api = {
   onPlayAudioStreamEnd: (callback) => {
     electron.ipcRenderer.on(IPC.PLAY_AUDIO_STREAM_END, () => callback());
   },
-  onStateChanged: (callback) => {
-    electron.ipcRenderer.on(IPC.STATE_CHANGED, (_event, state) => callback(state));
-  },
-  onStatusMessage: (callback) => {
-    electron.ipcRenderer.on(IPC.STATUS_MESSAGE, (_event, message) => callback(message));
-  },
   sendRecordingData: (data) => {
     electron.ipcRenderer.send(IPC.RECORDING_DATA, data);
   },

package/out/renderer/assets/index-CdX3ylbA.js

@@ -0,0 +1,209 @@
+const toggleOnUrl = "" + new URL("toggle_on-D9c1Kpa8.wav", import.meta.url).href;
+const toggleOffUrl = "" + new URL("toggle_off-DBeRrNFR.wav", import.meta.url).href;
+let mediaRecorder = null;
+let audioChunks = [];
+let audioContext = null;
+let pcmStream = null;
+let pcmSourceNode = null;
+let pcmProcessorNode = null;
+let pcmChunks = [];
+let pcmRecording = false;
+const WHISPER_SAMPLE_RATE = 16e3;
+function playSoundEffect(url) {
+  const ctx = audioContext ?? new AudioContext();
+  if (!audioContext) audioContext = ctx;
+  fetch(url).then((res) => res.arrayBuffer()).then((buf) => ctx.decodeAudioData(buf)).then((decoded) => {
+    const source = ctx.createBufferSource();
+    source.buffer = decoded;
+    const gain = ctx.createGain();
+    gain.gain.value = 2;
+    source.connect(gain);
+    gain.connect(ctx.destination);
+    source.start();
+  }).catch((err) => {
+    console.error("Failed to play sound effect:", err);
+  });
+}
+function downsample(buffer, sourceSampleRate, targetSampleRate) {
+  if (sourceSampleRate === targetSampleRate) return buffer;
+  const ratio = sourceSampleRate / targetSampleRate;
+  const newLength = Math.round(buffer.length / ratio);
+  const result = new Float32Array(newLength);
+  for (let i = 0; i < newLength; i++) {
+    const srcIndex = i * ratio;
+    const lo = Math.floor(srcIndex);
+    const hi = Math.min(lo + 1, buffer.length - 1);
+    const frac = srcIndex - lo;
+    result[i] = buffer[lo] * (1 - frac) + buffer[hi] * frac;
+  }
+  return result;
+}
+function startWebmRecording(stream) {
+  audioChunks = [];
+  mediaRecorder = new MediaRecorder(stream, {
+    mimeType: "audio/webm;codecs=opus"
+  });
+  mediaRecorder.ondataavailable = (event) => {
+    if (event.data.size > 0) {
+      audioChunks.push(event.data);
+    }
+  };
+  mediaRecorder.onstop = async () => {
+    stream.getTracks().forEach((track) => track.stop());
+    if (audioChunks.length === 0) {
+      window.piVoice.sendRecordingError("No audio data captured");
+      return;
+    }
+    const blob = new Blob(audioChunks, { type: "audio/webm" });
+    const arrayBuffer = await blob.arrayBuffer();
+    window.piVoice.sendRecordingData(arrayBuffer);
+  };
+  mediaRecorder.start(100);
+}
+function stopWebmRecording() {
+  if (mediaRecorder && mediaRecorder.state !== "inactive") {
+    mediaRecorder.stop();
+  }
+}
+function startPcmRecording(stream) {
+  const ctx = audioContext ?? new AudioContext();
+  if (!audioContext) audioContext = ctx;
+  pcmStream = stream;
+  pcmChunks = [];
+  pcmRecording = true;
+  pcmSourceNode = ctx.createMediaStreamSource(stream);
+  pcmProcessorNode = ctx.createScriptProcessor(4096, 1, 1);
+  pcmProcessorNode.onaudioprocess = (event) => {
+    if (!pcmRecording) return;
+    const input = event.inputBuffer.getChannelData(0);
+    pcmChunks.push(new Float32Array(input));
+  };
+  pcmSourceNode.connect(pcmProcessorNode);
+  pcmProcessorNode.connect(ctx.destination);
+}
+function stopPcmRecording() {
+  pcmRecording = false;
+  pcmProcessorNode?.disconnect();
+  pcmSourceNode?.disconnect();
+  pcmStream?.getTracks().forEach((track) => track.stop());
+  if (pcmChunks.length === 0) {
+    window.piVoice.sendRecordingError("No audio data captured");
+    pcmProcessorNode = null;
+    pcmSourceNode = null;
+    pcmStream = null;
+    return;
+  }
+  const totalLength = pcmChunks.reduce((sum, c) => sum + c.length, 0);
+  const fullBuffer = new Float32Array(totalLength);
+  let offset = 0;
+  for (const chunk of pcmChunks) {
+    fullBuffer.set(chunk, offset);
+    offset += chunk.length;
+  }
+  const sourceSampleRate = audioContext?.sampleRate ?? 48e3;
+  const resampled = downsample(fullBuffer, sourceSampleRate, WHISPER_SAMPLE_RATE);
+  window.piVoice.sendRecordingData(resampled.buffer);
+  pcmChunks = [];
+  pcmProcessorNode = null;
+  pcmSourceNode = null;
+  pcmStream = null;
+}
+let currentRecordingFormat = "webm";
+window.piVoice.onStartRecording(async (format) => {
+  playSoundEffect(toggleOnUrl);
+  currentRecordingFormat = format;
+  try {
+    const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
+    if (format === "pcm") {
+      startPcmRecording(stream);
+    } else {
+      startWebmRecording(stream);
+    }
+  } catch (err) {
+    const msg = err instanceof Error ? err.message : String(err);
+    window.piVoice.sendRecordingError(`Microphone access failed: ${msg}`);
+  }
+});
+window.piVoice.onStopRecording(() => {
+  playSoundEffect(toggleOffUrl);
+  if (currentRecordingFormat === "pcm") {
+    stopPcmRecording();
+  } else {
+    stopWebmRecording();
+  }
+});
+let streamSampleRate = 24e3;
+let streamChannels = 1;
+let streamBitsPerSample = 16;
+let streamNextPlayTime = 0;
+let streamActiveSources = 0;
+let streamEnded = false;
+function stopStreamPlayback() {
+  streamActiveSources = 0;
+  streamEnded = false;
+  streamNextPlayTime = 0;
+}
+window.piVoice.onPlayAudioStreamStart((meta) => {
+  try {
+    if (!audioContext) {
+      audioContext = new AudioContext();
+    }
+    stopStreamPlayback();
+    streamSampleRate = meta.sampleRate;
+    streamChannels = meta.channels;
+    streamBitsPerSample = meta.bitsPerSample;
+    streamNextPlayTime = 0;
+    streamEnded = false;
+  } catch (err) {
+    console.error("Stream start error:", err);
+  }
+});
+window.piVoice.onPlayAudioStreamChunk((pcmData) => {
+  try {
+    if (!audioContext) {
+      audioContext = new AudioContext();
+    }
+    const raw = pcmData instanceof ArrayBuffer ? pcmData : new Uint8Array(pcmData).buffer;
+    const bytesPerSample = streamBitsPerSample / 8;
+    const sampleCount = raw.byteLength / bytesPerSample / streamChannels;
+    if (sampleCount <= 0) return;
+    const audioBuffer = audioContext.createBuffer(
+      streamChannels,
+      sampleCount,
+      streamSampleRate
+    );
+    const view = new DataView(raw);
+    for (let ch = 0; ch < streamChannels; ch++) {
+      const channelData = audioBuffer.getChannelData(ch);
+      for (let i = 0; i < sampleCount; i++) {
+        const byteOffset = (i * streamChannels + ch) * bytesPerSample;
+        const int16 = view.getInt16(byteOffset, true);
+        channelData[i] = int16 / 32768;
+      }
+    }
+    const source = audioContext.createBufferSource();
+    source.buffer = audioBuffer;
+    source.connect(audioContext.destination);
+    const now = audioContext.currentTime;
+    if (streamNextPlayTime < now) {
+      streamNextPlayTime = now;
+    }
+    source.start(streamNextPlayTime);
+    streamNextPlayTime += audioBuffer.duration;
+    streamActiveSources++;
+    source.onended = () => {
+      streamActiveSources--;
+      if (streamEnded && streamActiveSources <= 0) {
+        window.piVoice.sendPlaybackDone();
+      }
+    };
+  } catch (err) {
+    console.error("Stream chunk playback error:", err);
+  }
+});
+window.piVoice.onPlayAudioStreamEnd(() => {
+  streamEnded = true;
+  if (streamActiveSources <= 0) {
+    window.piVoice.sendPlaybackDone();
+  }
+});

package/out/renderer/index.html

@@ -1,147 +1,10 @@
 <!DOCTYPE html>
-<html lang="ja">
+<html>
 <head>
   <meta charset="UTF-8">
-  <meta name="viewport" content="width=device-width, initial-scale=1.0">
-  <title>pi-voice</title>
-  <style>
-    * {
-      margin: 0;
-      padding: 0;
-      box-sizing: border-box;
-    }
-
-    body {
-      font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, sans-serif;
-      background: #1a1a2e;
-      color: #e0e0e0;
-      display: flex;
-      flex-direction: column;
-      align-items: center;
-      justify-content: center;
-      height: 100vh;
-      user-select: none;
-      -webkit-app-region: drag;
-      overflow: hidden;
-    }
-
-    .container {
-      display: flex;
-      flex-direction: column;
-      align-items: center;
-      gap: 24px;
-    }
-
-    .indicator {
-      width: 120px;
-      height: 120px;
-      border-radius: 50%;
-      background: #2a2a4a;
-      display: flex;
-      align-items: center;
-      justify-content: center;
-      transition: all 0.3s ease;
-      position: relative;
-    }
-
-    .indicator::after {
-      content: '';
-      position: absolute;
-      width: 100%;
-      height: 100%;
-      border-radius: 50%;
-      border: 3px solid transparent;
-      transition: all 0.3s ease;
-    }
-
-    /* State styles */
-    .state-idle .indicator {
-      background: #2a2a4a;
-    }
-
-    .state-recording .indicator {
-      background: #e74c3c;
-      box-shadow: 0 0 30px rgba(231, 76, 60, 0.5);
-      animation: pulse 1s ease-in-out infinite;
-    }
-
-    .state-transcribing .indicator {
-      background: #f39c12;
-      animation: spin 1.5s linear infinite;
-    }
-
-    .state-thinking .indicator {
-      background: #3498db;
-      animation: think 2s ease-in-out infinite;
-    }
-
-    .state-speaking .indicator {
-      background: #2ecc71;
-      animation: speak 0.5s ease-in-out infinite alternate;
-    }
-
-    .state-error .indicator {
-      background: #c0392b;
-    }
-
-    @keyframes pulse {
-      0%, 100% { transform: scale(1); }
-      50% { transform: scale(1.08); }
-    }
-
-    @keyframes spin {
-      from { transform: rotate(0deg); }
-      to { transform: rotate(360deg); }
-    }
-
-    @keyframes think {
-      0%, 100% { opacity: 1; }
-      50% { opacity: 0.5; }
-    }
-
-    @keyframes speak {
-      from { transform: scale(1); }
-      to { transform: scale(1.05); }
-    }
-
-    .icon {
-      font-size: 48px;
-      line-height: 1;
-    }
-
-    .state-label {
-      font-size: 16px;
-      font-weight: 600;
-      text-transform: uppercase;
-      letter-spacing: 2px;
-    }
-
-    .status-message {
-      font-size: 13px;
-      color: #888;
-      max-width: 300px;
-      text-align: center;
-      word-break: break-word;
-      min-height: 20px;
-    }
-
-    .hint {
-      font-size: 11px;
-      color: #555;
-      position: fixed;
-      bottom: 16px;
-    }
-  </style>
-  <script type="module" crossorigin src="./assets/index-dks-nI81.js"></script>
+  <title>pi-voice audio worker</title>
+  <script type="module" crossorigin src="./assets/index-CdX3ylbA.js"></script>
 </head>
 <body>
-  <div class="container" id="app">
-    <div class="indicator" id="indicator">
-      <span class="icon" id="icon">&#x23F8;</span>
-    </div>
-    <div class="state-label" id="stateLabel">IDLE</div>
-    <div class="status-message" id="statusMessage">Hold Fn to speak</div>
-  </div>
-  <div class="hint">Fn key: push-to-talk</div>
 </body>
 </html>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-voice",
-  "version": "0.1.0",
+  "version": "0.2.0",
   "description": "Voice interface for pi coding agent",
   "author": "Yuku Kotani",
   "license": "MIT",
@@ -11,8 +11,7 @@
   },
   "files": [
     "bin/",
-    "out/",
-    "build/"
+    "out/"
   ],
   "scripts": {
     "cli": "bun src/cli.ts",
@@ -23,14 +22,11 @@
     "build:cli": "bun build src/cli.ts --outdir out/cli --target node --format esm --external electron",
     "preview": "electron-vite preview",
     "prepack": "bun run build",
-    "dist": "bun run build && electron-builder --mac --config",
-    "dist:dir": "bun run build && electron-builder --mac --dir --config",
     "prepublish": "bun run build"
   },
   "devDependencies": {
     "@types/bun": "latest",
-    "electron-vite": "^3.1.0",
-    "electron-builder": "^26.0.0"
+    "electron-vite": "^3.1.0"
   },
   "peerDependencies": {
     "typescript": "^5"
@@ -38,12 +34,16 @@
   "dependencies": {
     "@google/genai": "^1.40.0",
     "@mariozechner/pi-coding-agent": "^0.52.7",
+    "@napi-rs/whisper": "^0.0.4",
     "electron": "^40.2.1",
-    "iohook-macos": "^1.2.1"
+    "openai": "^6.10.0",
+    "pino": "^10.3.1",
+    "uiohook-napi": "^1.5.4",
+    "zod": "^4.3.6"
   },
   "trustedDependencies": [
     "electron",
-    "iohook-macos",
+    "uiohook-napi",
     "protobufjs"
   ]
 }

package/build/entitlements.mac.plist

@@ -1,14 +0,0 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
-<plist version="1.0">
-<dict>
-    <key>com.apple.security.cs.allow-jit</key>
-    <true/>
-    <key>com.apple.security.cs.allow-unsigned-executable-memory</key>
-    <true/>
-    <key>com.apple.security.device.audio-input</key>
-    <true/>
-    <key>com.apple.security.automation.apple-events</key>
-    <true/>
-</dict>
-</plist>

package/out/renderer/assets/index-dks-nI81.js

@@ -1,162 +0,0 @@
-const toggleOnUrl = "" + new URL("toggle_on-D9c1Kpa8.wav", import.meta.url).href;
-const toggleOffUrl = "" + new URL("toggle_off-DBeRrNFR.wav", import.meta.url).href;
-document.getElementById("indicator");
-const icon = document.getElementById("icon");
-const stateLabel = document.getElementById("stateLabel");
-const statusMessage = document.getElementById("statusMessage");
-let mediaRecorder = null;
-let audioChunks = [];
-let audioContext = null;
-function playSoundEffect(url) {
-  const audio = new Audio(url);
-  audio.play().catch((err) => {
-    console.error("Failed to play sound effect:", err);
-  });
-}
-const stateConfig = {
-  idle: { icon: "⏸", label: "IDLE", defaultMessage: "Hold Fn to speak" },
-  recording: {
-    icon: "🔴",
-    label: "RECORDING",
-    defaultMessage: "Listening..."
-  },
-  transcribing: {
-    icon: "🔄",
-    label: "TRANSCRIBING",
-    defaultMessage: "Converting speech to text..."
-  },
-  thinking: {
-    icon: "🧠",
-    label: "THINKING",
-    defaultMessage: "pi is thinking..."
-  },
-  speaking: {
-    icon: "🔊",
-    label: "SPEAKING",
-    defaultMessage: "Playing response..."
-  },
-  error: { icon: "⚠", label: "ERROR", defaultMessage: "An error occurred" }
-};
-window.piVoice.onStateChanged((state) => {
-  document.body.className = "";
-  document.body.classList.add(`state-${state}`);
-  const config = stateConfig[state];
-  if (config) {
-    icon.textContent = config.icon;
-    stateLabel.textContent = config.label;
-    statusMessage.textContent = config.defaultMessage;
-  }
-});
-window.piVoice.onStatusMessage((message) => {
-  statusMessage.textContent = message;
-});
-window.piVoice.onStartRecording(async () => {
-  playSoundEffect(toggleOnUrl);
-  try {
-    const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
-    audioChunks = [];
-    mediaRecorder = new MediaRecorder(stream, {
-      mimeType: "audio/webm;codecs=opus"
-    });
-    mediaRecorder.ondataavailable = (event) => {
-      if (event.data.size > 0) {
-        audioChunks.push(event.data);
-      }
-    };
-    mediaRecorder.onstop = async () => {
-      stream.getTracks().forEach((track) => track.stop());
-      if (audioChunks.length === 0) {
-        window.piVoice.sendRecordingError("No audio data captured");
-        return;
-      }
-      const blob = new Blob(audioChunks, { type: "audio/webm" });
-      const arrayBuffer = await blob.arrayBuffer();
-      window.piVoice.sendRecordingData(arrayBuffer);
-    };
-    mediaRecorder.start(100);
-  } catch (err) {
-    const msg = err instanceof Error ? err.message : String(err);
-    window.piVoice.sendRecordingError(`Microphone access failed: ${msg}`);
-  }
-});
-window.piVoice.onStopRecording(() => {
-  playSoundEffect(toggleOffUrl);
-  if (mediaRecorder && mediaRecorder.state !== "inactive") {
-    mediaRecorder.stop();
-  }
-});
-let streamSampleRate = 24e3;
-let streamChannels = 1;
-let streamBitsPerSample = 16;
-let streamNextPlayTime = 0;
-let streamActiveSources = 0;
-let streamEnded = false;
-function stopStreamPlayback() {
-  streamActiveSources = 0;
-  streamEnded = false;
-  streamNextPlayTime = 0;
-}
-window.piVoice.onPlayAudioStreamStart((meta) => {
-  try {
-    if (!audioContext) {
-      audioContext = new AudioContext();
-    }
-    stopStreamPlayback();
-    streamSampleRate = meta.sampleRate;
-    streamChannels = meta.channels;
-    streamBitsPerSample = meta.bitsPerSample;
-    streamNextPlayTime = 0;
-    streamEnded = false;
-  } catch (err) {
-    console.error("Stream start error:", err);
-  }
-});
-window.piVoice.onPlayAudioStreamChunk((pcmData) => {
-  try {
-    if (!audioContext) {
-      audioContext = new AudioContext();
-    }
-    const raw = pcmData instanceof ArrayBuffer ? pcmData : new Uint8Array(pcmData).buffer;
-    const bytesPerSample = streamBitsPerSample / 8;
-    const sampleCount = raw.byteLength / bytesPerSample / streamChannels;
-    if (sampleCount <= 0) return;
-    const audioBuffer = audioContext.createBuffer(
-      streamChannels,
-      sampleCount,
-      streamSampleRate
-    );
-    const view = new DataView(raw);
-    for (let ch = 0; ch < streamChannels; ch++) {
-      const channelData = audioBuffer.getChannelData(ch);
-      for (let i = 0; i < sampleCount; i++) {
-        const byteOffset = (i * streamChannels + ch) * bytesPerSample;
-        const int16 = view.getInt16(byteOffset, true);
-        channelData[i] = int16 / 32768;
-      }
-    }
-    const source = audioContext.createBufferSource();
-    source.buffer = audioBuffer;
-    source.connect(audioContext.destination);
-    const now = audioContext.currentTime;
-    if (streamNextPlayTime < now) {
-      streamNextPlayTime = now;
-    }
-    source.start(streamNextPlayTime);
-    streamNextPlayTime += audioBuffer.duration;
-    streamActiveSources++;
-    source.onended = () => {
-      streamActiveSources--;
-      if (streamEnded && streamActiveSources <= 0) {
-        window.piVoice.sendPlaybackDone();
-      }
-    };
-  } catch (err) {
-    console.error("Stream chunk playback error:", err);
-  }
-});
-window.piVoice.onPlayAudioStreamEnd(() => {
-  streamEnded = true;
-  if (streamActiveSources <= 0) {
-    window.piVoice.sendPlaybackDone();
-  }
-});