opencode-interrupt-plugin 0.4.33 → 0.4.34

package/dist/config.d.ts

@@ -10,6 +10,7 @@ export interface PluginConfig {
     tts?: boolean;
     ttsVoice?: string;
     ttsRate?: string;
+    whisperModel?: string;
 }
 export interface ResolvedConfig {
     licenseKey?: string;
@@ -25,6 +26,7 @@ export interface ResolvedConfig {
     tts: boolean;
     ttsVoice: string;
     ttsRate: string;
+    whisperModel: string;
 }
 export declare const FREE_DEFAULTS: ResolvedConfig;
 export declare const SENSITIVITY_PRESETS: Record<string, Partial<ResolvedConfig>>;

package/dist/config.js

@@ -1,5 +1,21 @@
 const DEFAULT_TTS_VOICE = "en-US-AvaNeural";
 const DEFAULT_TTS_RATE = "+25%";
+const MODEL_NAMES = {
+    base: "ggml-base.bin",
+    small: "ggml-small.bin",
+    medium: "ggml-medium.bin",
+    large: "ggml-large-v3.bin",
+};
+function resolveModelPath(model) {
+    const home = process.env.HOME || "/tmp";
+    const envModel = process.env.WHISPER_MODEL;
+    if (envModel)
+        return envModel;
+    const filename = MODEL_NAMES[model];
+    if (filename)
+        return `${home}/.local/bin/${filename}`;
+    return model;
+}
 export const FREE_DEFAULTS = {
     isLicensed: false,
     micThreshold: 0.008,
@@ -13,6 +29,7 @@ export const FREE_DEFAULTS = {
     tts: true,
     ttsVoice: DEFAULT_TTS_VOICE,
     ttsRate: DEFAULT_TTS_RATE,
+    whisperModel: resolveModelPath("base"),
 };
 export const SENSITIVITY_PRESETS = {
     low: {
@@ -49,6 +66,9 @@ export function resolveConfig(userConfig, isLicensed) {
         tts: userConfig.tts ?? FREE_DEFAULTS.tts,
         ttsVoice: userConfig.ttsVoice ?? FREE_DEFAULTS.ttsVoice,
         ttsRate: userConfig.ttsRate ?? FREE_DEFAULTS.ttsRate,
+        whisperModel: userConfig.whisperModel
+            ? resolveModelPath(userConfig.whisperModel)
+            : FREE_DEFAULTS.whisperModel,
     };
     const preset = SENSITIVITY_PRESETS[base.sensitivity];
     if (!userConfig.timingWindowMs)

package/dist/index.js

@@ -12,7 +12,6 @@ import { readFileSync, existsSync, unlinkSync } from "node:fs";
 let activeSessionId = null;
 let pendingInterrupt = null;
 const RECORDING_FILE = "/tmp/interrupt-ptt.wav";
-const WHISPER_MODEL = process.env.WHISPER_MODEL || `${process.env.HOME}/.local/bin/ggml-base.bin`;
 let recordingProcess = null;
 let pttActive = false;
 function pttStartRecording() {
@@ -33,7 +32,7 @@ function pttStopRecording() {
     recordingProcess = null;
 }
 const TXT_FILE = "/tmp/interrupt-ptt.txt";
-async function transcribeLocal() {
+async function transcribeLocal(modelPath) {
     try {
         execSync("whisper --help", { stdio: "ignore", timeout: 3000 });
     }
@@ -45,7 +44,7 @@ async function transcribeLocal() {
     }
     catch { /* ignore */ }
     try {
-        execSync(`whisper -f "${RECORDING_FILE}" -m "${WHISPER_MODEL}" -otxt -of /tmp/interrupt-ptt`, { stdio: "ignore", timeout: 30000 });
+        execSync(`whisper -f "${RECORDING_FILE}" -m "${modelPath}" -otxt -of /tmp/interrupt-ptt`, { stdio: "ignore", timeout: 30000 });
     }
     catch {
         return null;
@@ -83,17 +82,25 @@ async function transcribeAPI() {
         return null;
     }
 }
-async function transcribeAndSend(sessionID, directory, api) {
+async function transcribeAndSend(sessionID, directory, api, modelPath) {
     pttStopRecording();
     if (!existsSync(RECORDING_FILE)) {
         api.ui.toast({ variant: "warning", title: "PTT", message: "No audio captured" });
         return;
     }
-    api.ui.toast({ variant: "info", title: "PTT", message: "⏳ Transcribing with Whisper..." });
+    const hasApiKey = !!process.env.OPENAI_API_KEY;
+    api.ui.toast({ variant: "info", title: "PTT", message: hasApiKey ? "⏳ Transcribing via OpenAI API..." : "⏳ Transcribing with Whisper..." });
     let text = null;
-    text = await transcribeLocal();
-    if (!text)
+    if (hasApiKey) {
         text = await transcribeAPI();
+        if (!text)
+            text = await transcribeLocal(modelPath);
+    }
+    else {
+        text = await transcribeLocal(modelPath);
+        if (!text)
+            text = await transcribeAPI();
+    }
     if (!text) {
         api.ui.toast({ variant: "error", title: "PTT", message: "❌ Install whisper: run scripts/install-whisper.sh, or set OPENAI_API_KEY" });
         try {
@@ -115,17 +122,25 @@ async function transcribeAndSend(sessionID, directory, api) {
     const preview = text.length > 80 ? text.slice(0, 77) + "..." : text;
     api.ui.toast({ variant: "success", title: "PTT", message: `✅ Sent: "${preview}"` });
 }
-async function transcribeAndSendV1(sessionID, client) {
+async function transcribeAndSendV1(sessionID, client, modelPath) {
     pttStopRecording();
     if (!existsSync(RECORDING_FILE)) {
         await client.tui.showToast({ body: { title: "PTT", message: "⚠️ No audio captured — try again", variant: "warning" } });
         return;
     }
-    await client.tui.showToast({ body: { title: "PTT", message: "⏳ Transcribing with Whisper...", variant: "info" } });
+    const hasApiKey = !!process.env.OPENAI_API_KEY;
+    await client.tui.showToast({ body: { title: "PTT", message: hasApiKey ? "⏳ Transcribing via OpenAI API..." : "⏳ Transcribing with Whisper...", variant: "info" } });
     let text = null;
-    text = await transcribeLocal();
-    if (!text)
+    if (hasApiKey) {
         text = await transcribeAPI();
+        if (!text)
+            text = await transcribeLocal(modelPath);
+    }
+    else {
+        text = await transcribeLocal(modelPath);
+        if (!text)
+            text = await transcribeAPI();
+    }
     if (!text) {
         await client.tui.showToast({ body: { title: "PTT", message: "❌ Install whisper: run scripts/install-whisper.sh, or set OPENAI_API_KEY", variant: "error", duration: 8000 } });
         try {
@@ -330,7 +345,7 @@ export const InterruptPlugin = (userConfig = {}) => {
                     const sessionID = cmdInput.sessionID;
                     if (pttActive) {
                         pttActive = false;
-                        await transcribeAndSendV1(sessionID, client);
+                        await transcribeAndSendV1(sessionID, client, config.whisperModel);
                     }
                     else {
                         pttActive = true;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-interrupt-plugin",
-  "version": "0.4.33",
+  "version": "0.4.34",
   "description": "Streaming TTS + voice interruption for OpenCode. Speaks responses as they arrive and detects when you talk over it.",
   "type": "module",
   "main": "./dist/index.js",