@lumiastream/wakeword 1.1.0 → 1.1.2

package/lib/list-devices.js

@@ -31,8 +31,9 @@ export function listAudioDevices() {
 			return;
 		}
 
-		// Windows: Parse available devices from SoX help output
-		const proc = spawn(soxPath, ["-t", "waveaudio", "-h"], {
+		// Windows: Use PowerShell to get audio devices
+		const psCommand = `Get-WmiObject Win32_SoundDevice | Select-Object -Property Name, DeviceID | ConvertTo-Json`;
+		const proc = spawn("powershell", ["-Command", psCommand], {
 			encoding: "utf8",
 			windowsHide: true,
 		});
@@ -51,69 +52,60 @@ export function listAudioDevices() {
 		proc.on("close", (code) => {
 			const devices = [];
 			
-			// Parse output for device list
-			const lines = (output + errorOutput).split("\n");
-			let inDeviceList = false;
-			
-			for (const line of lines) {
-				// Look for input device section
-				if (line.includes("Input:") || line.includes("input")) {
-					inDeviceList = true;
-					continue;
-				}
-				
-				// Stop at output section
-				if (line.includes("Output:") || line.includes("output")) {
-					inDeviceList = false;
-				}
-				
-				// Parse device entries (typically in format "0: Device Name")
-				if (inDeviceList) {
-					const match = line.match(/^\s*(\d+):\s*(.+)$/);
-					if (match) {
-						devices.push({
-							id: match[1],
-							name: match[2].trim(),
-						});
-					}
+			try {
+				// Try to parse PowerShell JSON output
+				if (output.trim()) {
+					const psDevices = JSON.parse(output);
+					const deviceArray = Array.isArray(psDevices) ? psDevices : [psDevices];
+					
+					deviceArray.forEach((device, index) => {
+						if (device && device.Name) {
+							devices.push({
+								id: index.toString(),
+								name: device.Name,
+							});
+						}
+					});
 				}
+			} catch (e) {
+				// PowerShell failed, try alternative method
 			}
 
+			// If PowerShell didn't work, try using Get-AudioDevice module if available
 			if (devices.length === 0) {
-				// Fallback: try to extract any numbered devices
-				const allMatches = (output + errorOutput).matchAll(/(\d+):\s*([^\n]+)/g);
-				for (const match of allMatches) {
-					const name = match[2].trim();
-					// Filter out obvious non-device entries
-					if (!name.includes("SoX") && !name.includes("Usage") && name.length > 0) {
-						devices.push({
-							id: match[1],
-							name: name,
-						});
-					}
+				// Provide numbered device options for manual selection
+				// SoX on Windows typically uses device numbers 0-9
+				for (let i = 0; i <= 5; i++) {
+					devices.push({
+						id: i.toString(),
+						name: `Audio Input Device ${i}${i === 0 ? ' (Default)' : ''}`,
+					});
 				}
-			}
-
-			// If still no devices found, provide default options
-			if (devices.length === 0) {
-				devices.push(
-					{ id: "0", name: "Device 0 (Default)" },
-					{ id: "1", name: "Device 1" },
-					{ id: "2", name: "Device 2" }
-				);
+				
+				console.log("\nNote: Unable to query actual device names.");
+				console.log("Showing generic device numbers (0-5).");
+				console.log("You may need to test each to find your microphone.\n");
 			}
 
 			resolve(devices);
 		});
 
 		proc.on("error", (err) => {
-			reject(err);
+			// If PowerShell is not available, provide default options
+			const devices = [];
+			for (let i = 0; i <= 5; i++) {
+				devices.push({
+					id: i.toString(),
+					name: `Audio Input Device ${i}${i === 0 ? ' (Default)' : ''}`,
+				});
+			}
+			resolve(devices);
 		});
 	});
 }
 
 // If run directly, list devices
-if (import.meta.url === `file://${process.argv[1]}`) {
+if (process.argv[1] && import.meta.url.endsWith(process.argv[1].replace(/\\/g, '/'))) {
 	listAudioDevices()
 		.then((devices) => {
 			console.log("Available audio input devices:");

package/lib/recorders/sox.js

@@ -1,53 +1,71 @@
 export default (options) => {
-	let cmd = "sox";
-
-	if (options.binPath) {
-		cmd = options.binPath;
-	}
-
-	let args = [
-		"--no-show-progress", // show no progress
-		"--rate",
-		options.sampleRate, // sample rate
-		"--channels",
-		options.channels, // channels
-		"--encoding",
-		"signed-integer", // sample encoding
-		"--bits",
-		"16", // precision (bits)
-		"--type",
-		options.audioType, // audio type
-		"-", // pipe
-	];
-
-	if (options.bufferSize) {
-		args.push("--buffer", options.bufferSize);
-	}
-
-	if (options.endOnSilence) {
-		args = args.concat([
-			"silence",
-			"1",
-			"0.1",
-			options.thresholdStart || options.threshold + "%",
-			"1",
-			options.silence,
-			options.thresholdEnd || options.threshold + "%",
-		]);
-	}
-
-	if (options.arguments) {
-		args = args.concat(options.arguments);
-	}
-
-	const spawnOptions = {};
-
-	if (options.device) {
-		args.unshift("-t", "waveaudio", options.device);
-		spawnOptions.env = { ...process.env, AUDIODEV: options.device };
-	} else {
-		args.unshift("--default-device");
-	}
-
-	return { cmd, args, spawnOptions };
+  let cmd = "sox";
+
+  if (options.binPath) {
+    cmd = options.binPath;
+  }
+
+  // Build common output/encoding args
+  let args = [
+    "--no-show-progress",
+    "--rate",
+    options.sampleRate,
+    "--channels",
+    options.channels,
+    "--encoding",
+    "signed-integer",
+    "--bits",
+    "16",
+    "--type",
+    options.audioType,
+    "-", // write to stdout
+  ];
+
+  if (options.bufferSize) {
+    args.push("--buffer", options.bufferSize);
+  }
+
+  if (options.endOnSilence) {
+    args = args.concat([
+      "silence",
+      "1",
+      "0.1",
+      options.thresholdStart || options.threshold + "%",
+      "1",
+      options.silence,
+      options.thresholdEnd || options.threshold + "%",
+    ]);
+  }
+
+  if (options.arguments) {
+    args = args.concat(options.arguments);
+  }
+
+  const spawnOptions = {};
+
+  // Prepend input spec based on platform
+  const platform = process.platform;
+  if (platform === "win32") {
+    const dev = options.device ?? "default";
+    args.unshift("-t", "waveaudio", dev);
+    // AUDIODEV sometimes respected on Windows; keep for compatibility
+    spawnOptions.env = { ...process.env, AUDIODEV: dev };
+  } else if (platform === "darwin") {
+    // CoreAudio input
+    if (options.device) {
+      args.unshift("-t", "coreaudio", options.device);
+    } else {
+      // Explicitly select CoreAudio default device for reliability
+      args.unshift("-t", "coreaudio", "default");
+    }
+  } else {
+    // Linux: ALSA default or specified
+    if (options.device) {
+      args.unshift("-t", "alsa", options.device);
+    } else {
+      args.unshift("-d");
+    }
+  }
+
+  return { cmd, args, spawnOptions };
 };

package/lib/test-devices.js

@@ -13,7 +13,9 @@ async function testDevice(deviceId) {
 	console.log("Speak into your microphone. Press Ctrl+C to stop.\n");
 
 	const voiceScript = join(here, "voice.js");
-	const proc = spawn("node", [voiceScript, undefined, deviceId], {
+	// Pass empty string for soxPath (will use default), then deviceId
+	const args = deviceId ? [voiceScript, "", deviceId] : [voiceScript];
+	const proc = spawn("node", args, {
 		stdio: ["pipe", "inherit", "inherit"],
 	});
 

package/lib/voice.js

@@ -25,7 +25,10 @@ const exeName = { win32: "sox.exe", darwin: "soxmac", linux: "soxlinux" }[
 
 /* Priority for sox path: argv[2] → fallback to sibling binaries/<exe> */
 /* Priority for device: argv[3] → env var → default */
-let soxPath = process.argv[2] || join(here, "..", "binaries", exeName);
+let soxPath = process.argv[2];
+if (!soxPath || soxPath === "") {
+	soxPath = join(here, "..", "binaries", exeName);
+}
 soxPath = unpacked(soxPath);
 
 // Parse device from argv[3] or environment variable
@@ -73,10 +76,18 @@ if (audioDevice !== null) {
 	// Windows: default to device 0 if not specified
 	recArgs.device = "0";
 	console.error("Using default Windows audio device: 0");
-	console.error("To specify a different device, use: AUDIO_DEVICE=<device_id> or pass as 3rd argument");
+	console.error(
+		"To specify a different device, use: AUDIO_DEVICE=<device_id> or pass as 3rd argument"
+	);
 }
 
 const mic = record.record(recArgs).stream();
+// Handle recorder (SoX) errors to avoid unhandled 'error' events
+mic.on("error", (err) => {
+	const msg = typeof err === "string" ? err : err?.message || String(err);
+	console.error(`[wakeword] audio stream error: ${msg}`);
+	process.exit(2);
+});
 // Define a confidence threshold for individual words.
 // You might need to adjust this value based on your specific use case.
 let WORD_CONFIDENCE_THRESHOLD = 0.7;

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lumiastream/wakeword",
-	"version": "1.1.0",
+	"version": "1.1.2",
 	"type": "module",
 	"main": "lib/index.js",
 	"files": [