shmakk 1.1.0 → 1.1.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "shmakk",
-  "version": "1.1.0",
+  "version": "1.1.1",
   "description": "AI-supervised terminal wrapper — command correction, tool-driven tasks, safety controls",
   "license": "MIT",
   "keywords": [

package/src/services/voice.js

@@ -17,7 +17,7 @@ const PAD_START_SEC = parseFloat(process.env.SHMAKK_VOICE_PAD_START_SEC || '0.3'
 // as noise/silence and never sent to Whisper. Tunable for noisy rooms.
 const MIN_RMS = parseFloat(process.env.SHMAKK_VOICE_MIN_RMS || '0.003');
 // Minimum captured speech duration in seconds (anything shorter is noise).
-const MIN_SPEECH_SEC = parseFloat(process.env.SHMAKK_VOICE_MIN_SPEECH_SEC || '0.8');
+const MIN_SPEECH_SEC = parseFloat(process.env.SHMAKK_VOICE_MIN_SPEECH_SEC || '0.5');
 
 // Track active TTS playback process so we can kill it on interrupt
 let _ttsProc = null;
@@ -206,7 +206,7 @@ function filterHallucination(text) {
   return text;
 }
 
-async function recordAndTranscribe({ language, maxDurationSec, onStart, onStop } = {}) {
+async function recordAndTranscribe({ language, maxDurationSec, onStart, onStop, suppressKillTts = false } = {}) {
   ensureAudioDir();
   const recorder = detectRecorder();
   if (!recorder) {
@@ -215,8 +215,9 @@ async function recordAndTranscribe({ language, maxDurationSec, onStart, onStop }
     );
   }
 
-  // Kill TTS so the AI stops talking when user starts speaking
-  _killTts();
+  // Kill TTS so the AI stops talking when user starts speaking.
+  // Suppressed in the interrupt-listener path where TTS is intentionally running.
+  if (!suppressKillTts) _killTts();
 
   const outFile = path.join(AUDIO_DIR, `voice-${Date.now()}.wav`);
   if (onStart) onStart();
@@ -235,9 +236,11 @@ async function recordAndTranscribe({ language, maxDurationSec, onStart, onStop }
   try {
     const { rms, durationSec } = audioStats(outFile);
     if (rms < MIN_RMS || durationSec < MIN_SPEECH_SEC) {
-      process.stderr.write(
-        `\r\x1b[90m[voice] skip (too quiet): rms=${rms.toFixed(4)} dur=${durationSec.toFixed(2)}s — tune with SHMAKK_VOICE_MIN_RMS\x1b[0m\n`,
-      );
+      if (rms > 0.0001) { // only log if there was actual audio — skip pure silence
+        process.stderr.write(
+          `\r\x1b[90m[voice] skip (too quiet): rms=${rms.toFixed(4)} dur=${durationSec.toFixed(2)}s — tune with SHMAKK_VOICE_MIN_RMS\x1b[0m\n`,
+        );
+      }
       cleanupFile(outFile);
       return '';
     }

package/src/session.js

@@ -35,6 +35,17 @@ const FLUSH_AFTER_BYTES = 8 * 1024;
 // limit are dropped from the front, preserving the most recent context).
 const HISTORY_MAX_ENTRIES = 30;
 
+// Kitty terminal sends \x1b[99;5u instead of \x03 for Ctrl+C.
+// Returns the byte index of the first Ctrl+C (either form), or -1.
+const KITTY_CTRL_C = Buffer.from([0x1b, 0x5b, 0x39, 0x39, 0x3b, 0x35, 0x75]); // \x1b[99;5u
+function findCtrlC(data) {
+  for (let i = 0; i < data.length; i++) {
+    if (data[i] === 0x03) return i;
+    if (data[i] === 0x1b && data.slice(i, i + KITTY_CTRL_C.length).equals(KITTY_CTRL_C)) return i;
+  }
+  return -1;
+}
+
 function isAbortError(e) {
   return e && (e.name === 'AbortError' || /aborted/i.test(String(e.message || '')));
 }
@@ -117,20 +128,14 @@ async function runOneSession(opts, registerSession) {
   audit.append({ kind: 'session-start', workspace: cwd, pinnedWorkspace, review: !!opts.review, pid: process.pid });
 
   // ── Global Ctrl+C handler (persistent bottom-of-stack) ──
-  // Sits below any AI-task handler. When no AI is running, this is the
-  // top handler and intercepts Ctrl+C to:
-  // 1. Stop TTS playback if active
-  // 2. Kill voice recording if in progress
-  // 3. On a second Ctrl+C within 2s while --sts is active, exit the
-  //    always-on voice loop so the user is left at a normal shell prompt.
-  // If none apply, pass through to the child shell.
-  // Ctrl+C = shut up. Kills TTS, recorder, and voice loop. Always.
-  // Ctrl+D exits the shell as normal (we never touch it).
+  // Ctrl+C = shut up. Kills TTS, recorder, and voice loop immediately.
+  // Ctrl+D exits the shell as normal (we never intercept it).
   session.captureStdin((data) => {
-    for (let i = 0; i < data.length; i++) {
-      if (data[i] === 0x03 && (opts.tts || opts.stt || opts.sts)) {
+    if (opts.tts || opts.stt || opts.sts) {
+      const cut = findCtrlC(data);
+      if (cut !== -1) {
         try { fullVoiceTeardown(); } catch {}
-        if (i > 0) session.childWrite(data.slice(0, i));
+        if (cut > 0) session.childWrite(data.slice(0, cut));
         session.childWrite('\r');
         return;
       }
@@ -175,10 +180,7 @@ async function runOneSession(opts, registerSession) {
     const ctrl = new AbortController();
     setMaxListeners(0, ctrl.signal);
     const release = session.captureStdin((data) => {
-      let cut = -1;
-      for (let i = 0; i < data.length; i++) {
-        if (data[i] === 0x03) { cut = i; break; }
-      }
+      const cut = findCtrlC(data);
       if (cut === -1) {
         session.childWrite(data);
         return;
@@ -250,6 +252,7 @@ async function runOneSession(opts, registerSession) {
             history,
             profile: opts.profile,
             colors: colorsEnabled,
+            voiceMode: true,
           });
           history = trimHistory(updated || history);
           if ((opts.tts || opts.sts) && updated && updated.length) {
@@ -272,7 +275,30 @@ async function runOneSession(opts, registerSession) {
                   if (session._stsFlags) session._stsFlags.setTtsSpeaking(false);
                   if (err && opts.debug) process.stderr.write(`[shmakk] tts: ${err.message}\n`);
                 };
-                tts.speak(reply, { voice: ttsVoice }).then(() => settle()).catch(settle);
+                // Parallel interrupt listener — lets user say "stop" to cut TTS.
+                // suppressKillTts=true so recording alongside TTS doesn't immediately kill it.
+                // Loop is gated on myGen so it stops the moment settle() fires.
+                const STOP_WORDS = new Set(['stop', 'quiet', 'shut up', 'silence', 'enough', 'cancel']);
+                let interruptListening = true;
+                const listenForInterrupt = async () => {
+                  const vs = getVoiceService();
+                  while (interruptListening && session._ttsGen === myGen) {
+                    try {
+                      const heard = await vs.recordAndTranscribe({ maxDurationSec: 2, suppressKillTts: true });
+                      if (!heard) continue;
+                      if (STOP_WORDS.has(heard.toLowerCase().trim().replace(/[.!?]$/, ''))) {
+                        try { fullVoiceTeardown(); } catch {}
+                        break;
+                      }
+                    } catch { break; }
+                  }
+                };
+                listenForInterrupt().catch(() => {});
+                const settleAndStop = (err) => {
+                  interruptListening = false; // stop interrupt loop before unpausing voice loop
+                  settle(err);
+                };
+                tts.speak(reply, { voice: ttsVoice }).then(() => settleAndStop()).catch(settleAndStop);
               }
             }
           }
@@ -372,7 +398,7 @@ async function runOneSession(opts, registerSession) {
         let recordingDone = false;
         const release = session.captureStdin((data) => {
           for (let i = 0; i < data.length; i++) {
-            if (data[i] === 0x03 || data[i] === 0x0f) {
+            if (data[i] === 0x03 || data[i] === 0x0f || findCtrlC(data) !== -1) {
               recordingDone = true;
               // Kill the recorder process immediately
               try { vs._killRecorder(); } catch {}

package/src/system-prompt.js

@@ -10,6 +10,7 @@ function buildSystemPrompt({
   activeSkillText,
   maxDiscoveryCallsPerRound,
   runtimeProfile,
+  voiceMode = false,
 }) {
   return `You are an expert AI coding assistant running inside shmakk.
 
@@ -255,6 +256,7 @@ Otherwise output only:
 {"shmakk_actions":[{"tool":"tool_name","args":{...}}]}
 ${indexHint}
 ${activeSkillText ? `\n\n${activeSkillText}` : ''}
+${voiceMode ? `\n\nVOICE MODE: The user is speaking to you. Keep every reply under 2 sentences. No bullet points, no markdown, no code blocks unless explicitly asked. Speak like a human, not a document.` : ''}
 `;
 }