speech-opencode 1.0.0 → 1.1.1

package/README.md

@@ -2,7 +2,7 @@
 
 Voice input plugin for [OpenCode](https://opencode.ai) using OpenAI Whisper.
 
-Record audio from your microphone and transcribe it to text using OpenAI's Whisper API.
+Record audio from your microphone and transcribe it to text using OpenAI's Whisper API. **Recording automatically stops when you stop talking** - no need to specify a duration!
 
 ## Installation
 
@@ -26,33 +26,32 @@ export OPENAI_API_KEY=your-api-key
 
 ### Audio Recording Tools
 
-**Linux (PulseAudio/PipeWire):**
+**sox** is required for audio recording with silence detection:
+
 ```bash
+# macOS
+brew install sox
+
 # Ubuntu/Debian
-sudo apt install pulseaudio-utils
+sudo apt install sox
 
 # Fedora
-sudo dnf install pulseaudio-utils
+sudo dnf install sox
 
 # Arch
-sudo pacman -S pulseaudio-utils
-```
-
-**macOS:**
-```bash
-brew install sox
+sudo pacman -S sox
 ```
 
 ## Usage
 
-Once installed, OpenCode will have access to a `voice` tool. You can ask OpenCode to use it:
+Once installed, OpenCode will have access to a `voice` tool. Just ask OpenCode:
 
 - "Listen to my voice"
 - "Record what I say"
 - "Use voice input"
-- "Transcribe my speech for 10 seconds"
+- "voice"
 
-The tool accepts an optional `duration` parameter (default: 5 seconds, max: 60 seconds).
+**Recording automatically stops after 7 seconds of silence**, so just speak naturally and pause when you're done.
 
 ## Configuration
 
@@ -66,11 +65,11 @@ export default VoicePlugin({
   // Optional: specify language (auto-detects if not set)
   language: "en",
   
-  // Optional: default recording duration in seconds
-  defaultDuration: 5,
+  // Optional: seconds of silence before stopping (default 7)
+  silenceDuration: 7,
   
-  // Optional: maximum recording duration in seconds
-  maxDuration: 60,
+  // Optional: maximum recording time as safety timeout (default 300 = 5 min)
+  maxDuration: 300,
   
   // Optional: override API key (defaults to OPENAI_API_KEY env var)
   apiKey: process.env.MY_OPENAI_KEY,
@@ -92,9 +91,10 @@ Leave `language` unset for automatic detection.
 
 ## How It Works
 
-1. Records audio from your default microphone using system tools
-2. Sends the audio to OpenAI's Whisper API for transcription
-3. Returns the transcribed text to OpenCode
+1. Starts recording from your microphone when you begin speaking
+2. Automatically stops after detecting 7 seconds of silence
+3. Sends the audio to OpenAI's Whisper API for transcription
+4. Returns the transcribed text to OpenCode
 
 ## Troubleshooting
 
@@ -103,8 +103,12 @@ Leave `language` unset for automatic detection.
 - Verify the correct input device is selected in your system settings
 - On Linux, use `pavucontrol` to check input sources
 
+### Recording doesn't stop
+- Make sure you pause speaking for at least 7 seconds
+- Check that background noise isn't being detected as speech
+
 ### Recording fails
-- Ensure you have the required audio tools installed
+- Ensure sox is installed: `which rec`
 - Check that your microphone permissions are granted
 
 ## License

package/dist/index.d.ts

@@ -4,9 +4,9 @@ export interface VoicePluginOptions {
     apiKey?: string;
     /** Language code for transcription (e.g., "en", "es", "fr"). Auto-detects if not specified */
     language?: string;
-    /** Default recording duration in seconds */
-    defaultDuration?: number;
-    /** Maximum allowed recording duration in seconds */
+    /** Seconds of silence before stopping recording (default 7) */
+    silenceDuration?: number;
+    /** Maximum recording duration in seconds as a safety timeout (default 300 = 5 minutes) */
     maxDuration?: number;
 }
 /**

package/dist/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,KAAK,MAAM,EAAQ,MAAM,qBAAqB,CAAA;AA6LvD,MAAM,WAAW,kBAAkB;IACjC,yDAAyD;IACzD,MAAM,CAAC,EAAE,MAAM,CAAA;IACf,8FAA8F;IAC9F,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,4CAA4C;IAC5C,eAAe,CAAC,EAAE,MAAM,CAAA;IACxB,oDAAoD;IACpD,WAAW,CAAC,EAAE,MAAM,CAAA;CACrB;AAED;;;;;;;;;;;;;;;;;;;;GAoBG;AACH,eAAO,MAAM,WAAW,GACrB,UAAS,kBAAuB,KAAG,MAoEnC,CAAA;;AAGH,wBAA4B"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,KAAK,MAAM,EAAQ,MAAM,qBAAqB,CAAA;AAwGvD,MAAM,WAAW,kBAAkB;IACjC,yDAAyD;IACzD,MAAM,CAAC,EAAE,MAAM,CAAA;IACf,8FAA8F;IAC9F,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,+DAA+D;IAC/D,eAAe,CAAC,EAAE,MAAM,CAAA;IACxB,0FAA0F;IAC1F,WAAW,CAAC,EAAE,MAAM,CAAA;CACrB;AAED;;;;;;;;;;;;;;;;;;;;GAoBG;AACH,eAAO,MAAM,WAAW,GACrB,UAAS,kBAAuB,KAAG,MA4DnC,CAAA;;AAGH,wBAA4B"}

package/dist/index.js

@@ -5,127 +5,56 @@ import { unlinkSync, readFileSync } from "fs";
 import { tmpdir } from "os";
 import { join } from "path";
 /**
- * Gets the first available non-monitor, non-bluetooth audio input source
- * Works with PulseAudio and PipeWire on Linux
+ * Records audio from the microphone with automatic silence detection.
+ * Recording stops after the specified silence duration.
+ * Uses sox on both Linux and macOS for silence detection.
+ *
+ * @param maxDurationSeconds - Maximum recording time (safety timeout)
+ * @param silenceDuration - Seconds of silence before stopping (default 7)
  */
-async function getDefaultInputDevice() {
-    return new Promise((resolve) => {
-        const pactl = spawn("pactl", ["list", "sources", "short"]);
-        let output = "";
-        pactl.stdout.on("data", (data) => {
-            output += data.toString();
-        });
-        pactl.on("close", () => {
-            const lines = output.trim().split("\n");
-            for (const line of lines) {
-                const parts = line.split("\t");
-                if (parts.length >= 2) {
-                    const name = parts[1];
-                    // Skip monitor sources and bluetooth (prefer hardware input)
-                    if (!name.includes(".monitor") && !name.includes("bluez")) {
-                        resolve(name);
-                        return;
-                    }
-                }
-            }
-            resolve(null);
-        });
-        pactl.on("error", () => resolve(null));
-    });
+async function recordAudio(maxDurationSeconds = 300, silenceDuration = 7) {
+    const tempFile = join(tmpdir(), `opencode-voice-${Date.now()}.wav`);
+    // Use sox with silence detection on all platforms
+    return recordWithSilenceDetection(tempFile, maxDurationSeconds, silenceDuration);
 }
 /**
- * Records audio from the microphone
- * - Linux: Uses parecord (PulseAudio/PipeWire) or arecord (ALSA)
- * - macOS: Uses sox (rec command)
+ * Records audio using sox with silence detection.
+ * Recording automatically stops after detecting silence.
+ *
+ * Sox silence syntax: silence [above_periods] [duration] [threshold] [below_periods] [duration] [threshold]
+ * - above_periods 1: need 1 period of audio above threshold to start
+ * - 0.1 3%: audio must be above 3% for 0.1s to count as speech start
+ * - below_periods 1: need 1 period below threshold to stop
+ * - silenceDuration 3%: stop after silenceDuration seconds below 3%
  */
-async function recordAudio(durationSeconds = 5) {
-    const tempFile = join(tmpdir(), `opencode-voice-${Date.now()}.wav`);
-    const platform = process.platform;
-    if (platform === "darwin") {
-        // macOS: use sox
-        return recordWithSox(tempFile, durationSeconds);
-    }
-    else {
-        // Linux: use parecord or arecord
-        return recordWithPulseAudio(tempFile, durationSeconds);
-    }
-}
-async function recordWithSox(tempFile, durationSeconds) {
+async function recordWithSilenceDetection(tempFile, maxDurationSeconds, silenceDuration) {
     return new Promise((resolve, reject) => {
-        const recorder = spawn("rec", [
+        // Use timeout to enforce max duration, sox for silence detection
+        const recorder = spawn("timeout", [
+            maxDurationSeconds.toString(),
+            "rec",
             "-q",
-            "-r",
-            "16000",
-            "-c",
-            "1",
-            "-b",
-            "16",
+            "-r", "16000",
+            "-c", "1",
+            "-b", "16",
             tempFile,
-            "trim",
-            "0",
-            durationSeconds.toString(),
+            "silence",
+            "1", "0.1", "3%", // Start recording when speech detected (above 3% for 0.1s)
+            "1", `${silenceDuration}.0`, "3%", // Stop after silenceDuration seconds of silence (below 3%)
         ]);
         let errorOutput = "";
         recorder.stderr.on("data", (data) => {
             errorOutput += data.toString();
         });
         recorder.on("error", () => {
-            reject(new Error("sox not found. Please install it:\n" + "  - macOS: brew install sox"));
-        });
-        recorder.on("close", (code) => {
-            if (code === 0) {
-                resolve(tempFile);
-            }
-            else {
-                reject(new Error(`Recording failed: ${errorOutput}`));
-            }
-        });
-    });
-}
-async function recordWithPulseAudio(tempFile, durationSeconds) {
-    const inputDevice = await getDefaultInputDevice();
-    return new Promise((resolve, reject) => {
-        const args = [(durationSeconds + 1).toString(), "parecord"];
-        if (inputDevice) {
-            args.push(`--device=${inputDevice}`);
-        }
-        args.push("--file-format=wav", tempFile);
-        const recorder = spawn("timeout", args);
-        let errorOutput = "";
-        recorder.stderr.on("data", (data) => {
-            errorOutput += data.toString();
-        });
-        recorder.on("error", () => {
-            // Fallback to arecord
-            const arecord = spawn("arecord", [
-                "-q",
-                "-f",
-                "S16_LE",
-                "-r",
-                "16000",
-                "-c",
-                "1",
-                "-d",
-                durationSeconds.toString(),
-                tempFile,
-            ]);
-            arecord.on("error", () => {
-                reject(new Error("No audio recorder found. Please install:\n" +
-                    "  - Ubuntu/Debian: sudo apt install pulseaudio-utils\n" +
-                    "  - Fedora: sudo dnf install pulseaudio-utils\n" +
-                    "  - Arch: sudo pacman -S pulseaudio-utils"));
-            });
-            arecord.on("close", (code) => {
-                if (code === 0) {
-                    resolve(tempFile);
-                }
-                else {
-                    reject(new Error(`arecord failed with code ${code}`));
-                }
-            });
+            reject(new Error("sox not found. Please install it:\n" +
+                "  - macOS: brew install sox\n" +
+                "  - Ubuntu/Debian: sudo apt install sox\n" +
+                "  - Fedora: sudo dnf install sox\n" +
+                "  - Arch: sudo pacman -S sox"));
         });
         recorder.on("close", (code) => {
-            // timeout returns 124 when it kills the process, which is expected
+            // code 0 = normal exit, 124 = timeout killed it (max duration reached)
             if (code === 0 || code === 124) {
                 resolve(tempFile);
             }
@@ -171,7 +100,7 @@ async function transcribeAudio(audioFilePath, apiKey, language) {
  * ```
  */
 export const VoicePlugin = (options = {}) => async (ctx) => {
-    const { apiKey = process.env.OPENAI_API_KEY, language, defaultDuration = 5, maxDuration = 60, } = options;
+    const { apiKey = process.env.OPENAI_API_KEY, language, silenceDuration = 7, maxDuration = 300, } = options;
     if (!apiKey) {
         console.warn("[Voice Plugin] Warning: OPENAI_API_KEY not set. Voice transcription will fail.");
     }
@@ -180,21 +109,15 @@ export const VoicePlugin = (options = {}) => async (ctx) => {
             voice: tool({
                 description: "Records audio from the user's microphone and transcribes it using OpenAI Whisper. " +
                     "Use this tool when the user wants to provide input via voice or speech. " +
-                    `The tool will record for the specified duration (default ${defaultDuration} seconds) and return the transcribed text.`,
-                args: {
-                    duration: tool.schema
-                        .number()
-                        .optional()
-                        .describe(`Recording duration in seconds. Default is ${defaultDuration} seconds. Max is ${maxDuration} seconds.`),
-                },
-                async execute(args) {
+                    "Recording automatically stops after detecting silence, so the user can speak naturally without specifying a duration.",
+                args: {},
+                async execute() {
                     if (!apiKey) {
                         return "Error: OPENAI_API_KEY environment variable is not set. Please set it to use voice transcription.";
                     }
-                    const duration = Math.min(args.duration || defaultDuration, maxDuration);
                     let audioFile = null;
                     try {
-                        audioFile = await recordAudio(duration);
+                        audioFile = await recordAudio(maxDuration, silenceDuration);
                         const transcription = await transcribeAudio(audioFile, apiKey, language);
                         if (!transcription || transcription.trim() === "") {
                             return "No speech detected. Please try again and speak clearly into your microphone.";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "speech-opencode",
-  "version": "1.0.0",
+  "version": "1.1.1",
   "description": "Voice input plugin for OpenCode using OpenAI Whisper",
   "keywords": [
     "opencode",