@codexstar/pi-listen 1.0.14 → 1.0.15

package/extensions/voice.ts

@@ -636,23 +636,31 @@ export default function (pi: ExtensionAPI) {
 	/** Animated recording indicator with live waveform */
 	function showRecordingWidget(target: "editor" | "btw") {
 		if (!ctx?.hasUI) return;
-		let frame = 0;
-		const waveChars = ["▁", "▂", "▃", "▅", "▆", "▇", "▆", "▅", "▃", "▂"];
 
-		// Animate the widget every 200ms
+		// Store initial state — once live transcription arrives,
+		// updateLiveTranscriptWidget takes over and we stop the animation.
+		(showRecordingWidget as any)._target = target;
+		(showRecordingWidget as any)._frame = 0;
+		(showRecordingWidget as any)._hasTranscript = false;
+
+		// Animate the widget every 300ms (only while no transcript is showing)
 		const animTimer = setInterval(() => {
-			frame++;
-			if (ctx?.hasUI) ctx.ui.setWidget("voice-recording", undefined); // force re-render
-			showRecordingWidgetFrame(target, frame, waveChars);
-		}, 200);
+			// Stop animating once live transcript takes over
+			if ((showRecordingWidget as any)?._hasTranscript) return;
+
+			(showRecordingWidget as any)._frame = ((showRecordingWidget as any)._frame || 0) + 1;
+			showRecordingWidgetFrame(target, (showRecordingWidget as any)._frame);
+		}, 300);
 
 		// Store the timer so we can clean it up
 		(showRecordingWidget as any)._animTimer = animTimer;
 
-		showRecordingWidgetFrame(target, frame, waveChars);
+		showRecordingWidgetFrame(target, 0);
 	}
 
-	function showRecordingWidgetFrame(target: "editor" | "btw", frame: number, waveChars: string[]) {
+	const waveChars = ["▁", "▂", "▃", "▅", "▆", "▇", "▆", "▅", "▃", "▂"];
+
+	function showRecordingWidgetFrame(target: "editor" | "btw", frame: number) {
 		if (!ctx?.hasUI) return;
 		ctx.ui.setWidget("voice-recording", (tui, theme) => {
 			return {
@@ -691,7 +699,7 @@ export default function (pi: ExtensionAPI) {
 						? theme.fg("dim", "  Press Ctrl+Shift+B to stop")
 						: kittyReleaseDetected
 							? theme.fg("dim", "  Release SPACE to finalize")
-							: theme.fg("dim", "  Press SPACE again to stop");
+							: theme.fg("dim", "  Press Ctrl+Shift+V to stop");
 
 					const lines = [
 						topBorder,
@@ -717,6 +725,10 @@ export default function (pi: ExtensionAPI) {
 	function updateLiveTranscriptWidget(interim: string, finals: string[]) {
 		if (!ctx?.hasUI) return;
 
+		// Stop the recording animation — live transcript takes over
+		(showRecordingWidget as any)._hasTranscript = true;
+		stopRecordingWidgetAnimation();
+
 		const finalized = finals.join(" ");
 		const displayText = finalized + (interim ? (finalized ? " " : "") + interim : "");
 
@@ -745,7 +757,7 @@ export default function (pi: ExtensionAPI) {
 					const titleLine = `  ${dot} ${label}  ${timeStyled}`;
 					const hint = kittyReleaseDetected
 						? theme.fg("dim", "  Release SPACE to finalize")
-						: theme.fg("dim", "  Press SPACE again to stop");
+						: theme.fg("dim", "  Press Ctrl+Shift+V to stop");
 
 					const lines = [topBorder, side(titleLine)];
 
@@ -1005,13 +1017,19 @@ export default function (pi: ExtensionAPI) {
 	// For Kitty protocol terminals: hold → wait threshold → activate →
 	//   release → stop recording. True hold-to-talk.
 	// For non-Kitty terminals: hold → wait threshold → activate →
-	//   press SPACE again → stop recording. Toggle after activation.
+	//   Ctrl+Shift+V or /voice stop to end recording.
+	//
+	// KEY INSIGHT: In non-Kitty terminals, holding a key generates
+	// rapid press events (key-repeat). We CANNOT use "second space press
+	// = stop" because repeats arrive while holding. Instead, non-Kitty
+	// users must use Ctrl+Shift+V to stop.
 
 	const HOLD_THRESHOLD_MS = 500; // minimum hold time before voice activates
 	let kittyReleaseDetected = false;
 	let spaceDownTime: number | null = null; // timestamp when SPACE was first pressed
 	let holdActivationTimer: ReturnType<typeof setTimeout> | null = null;
 	let spaceConsumed = false; // whether we've committed to voice (past threshold)
+	let lastSpacePressTime = 0; // debounce rapid space presses from key-repeat
 
 	function clearHoldTimer() {
 		if (holdActivationTimer) {
@@ -1054,7 +1072,7 @@ export default function (pi: ExtensionAPI) {
 					}
 
 					// Released after threshold → stop recording (true hold-to-talk)
-					if (spaceConsumed && isHolding && voiceState === "recording") {
+					if (spaceConsumed && voiceState === "recording") {
 						isHolding = false;
 						spaceConsumed = false;
 						spaceDownTime = null;
@@ -1067,32 +1085,42 @@ export default function (pi: ExtensionAPI) {
 					return undefined;
 				}
 
-				// ── Kitty key-repeat: suppress while holding past threshold ──
+				// ── Kitty key-repeat: ALWAYS suppress while holding/recording ──
 				if (isKeyRepeat(data)) {
-					if (spaceConsumed || isHolding) return { consume: true };
+					if (spaceDownTime || spaceConsumed || isHolding || voiceState === "recording") {
+						return { consume: true };
+					}
 					return undefined;
 				}
 
-				// === Key PRESS ===
+				// === Key PRESS (initial press only) ===
 
-				// If already recording (toggle mode for non-Kitty) → stop
-				if (voiceState === "recording" && spaceConsumed) {
-					isHolding = false;
-					spaceConsumed = false;
-					spaceDownTime = null;
-					clearHoldTimer();
-					stopVoiceRecording("editor");
-					return { consume: true };
+				const now = Date.now();
+
+				// Debounce: ignore rapid presses within 100ms (terminal key-repeat
+				// generates press events in non-Kitty terminals since there's no
+				// key-repeat flag — they all look like fresh presses)
+				if (now - lastSpacePressTime < 100) {
+					lastSpacePressTime = now;
+					return { consume: true }; // suppress repeat
 				}
+				lastSpacePressTime = now;
 
 				// If transcribing → ignore
 				if (voiceState === "transcribing") {
 					return { consume: true };
 				}
 
+				// If already recording: In Kitty mode, release handles stop.
+				// In non-Kitty, we can't safely detect "real second press" vs
+				// key-repeat. Use Ctrl+Shift+V instead. Just consume.
+				if (voiceState === "recording") {
+					return { consume: true };
+				}
+
 				// Idle → start the hold timer
 				if (voiceState === "idle" && !spaceDownTime) {
-					spaceDownTime = Date.now();
+					spaceDownTime = now;
 					spaceConsumed = false;
 
 					// Show a subtle "preparing" indicator
@@ -1326,9 +1354,19 @@ export default function (pi: ExtensionAPI) {
 				return;
 			}
 			if (voiceState === "idle") {
-				await startVoiceRecording("editor");
+				// Direct start — bypass hold threshold
+				spaceConsumed = true;
+				isHolding = true;
+				const ok = await startVoiceRecording("editor");
+				if (!ok) {
+					isHolding = false;
+					spaceConsumed = false;
+				}
 			} else if (voiceState === "recording") {
 				isHolding = false;
+				spaceConsumed = false;
+				spaceDownTime = null;
+				clearHoldTimer();
 				await stopVoiceRecording("editor");
 			}
 		},

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@codexstar/pi-listen",
-  "version": "1.0.14",
+  "version": "1.0.15",
   "description": "Voice input, first-run onboarding, and side-channel BTW conversations for Pi",
   "type": "module",
   "keywords": [