fix(tts): stopped queued TTS playback on Esc after stream end

Once the assistant reply stops streaming, `vocalizer.clear()` was only invoked from the aborted-stream cascade in EventController. Escaping after the model finished fell through InputController to the empty-editor double-Esc gesture while StreamingAudioPlayer kept draining buffered Kokoro PCM.

Add `Vocalizer.isSpeaking()` (true while any live player, stream handle, or in-flight abort is around) and consult it in the Esc handler before the double-Esc branch: if speech is still audible, a single Esc calls `vocalizer.clear()` and resets `lastEscapeTime` so tree/branch stays reachable via the next press.

Fixes #4521

packages/coding-agent/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+### Fixed
+
+- Fixed Esc no longer stopping TTS audio playback once the assistant reply had finished streaming; queued Kokoro (and remote) audio kept reading past the response's end because `vocalizer.clear()` only fired from the aborted-stream cascade. Esc now silences a still-audible vocalizer as its first action (ahead of the `tree`/`branch` double-Esc gesture); a second Esc keeps its previous behavior. ([#4521](https://github.com/can1357/oh-my-pi/issues/4521))
+
 ## [16.3.6] - 2026-07-04
 
 ### Changed

packages/coding-agent/src/modes/controllers/input-controller.ts

@@ -23,6 +23,7 @@ import { isLowSignalTitleInput } from "../../tiny/text";
 import { tinyTitleClient } from "../../tiny/title-client";
 import type { TinyTitleProgressEvent } from "../../tiny/title-protocol";
 import { shortenPath, TRUNCATE_LENGTHS, truncateToWidth } from "../../tools/render-utils";
+import { vocalizer } from "../../tts/vocalizer";
 import {
 	copyToClipboard,
 	readImageFromClipboard,
@@ -390,6 +391,13 @@ export class InputController {
 				// Esc must not destroy an in-progress draft; it only disarms a previous empty-editor Esc.
 				this.ctx.lastEscapeTime = 0;
 				this.#clearStreamingEscapeArm();
+			} else if (vocalizer.isSpeaking()) {
+				// TTS buffers seconds of PCM past the streaming abort, so an Esc
+				// arriving after the model stopped would otherwise fall through to
+				// the double-Esc gesture while Kokoro reads on. Silence first;
+				// tree/branch stays reachable via a second Esc.
+				vocalizer.clear();
+				this.ctx.lastEscapeTime = 0;
 			} else {
 				// Double-interrupt with empty editor triggers /tree, /branch, or nothing based on setting
 				const action = settings.get("doubleEscapeAction");

packages/coding-agent/src/tts/vocalizer.ts

@@ -195,6 +195,15 @@ export class Vocalizer {
 		this.#livePlayers.clear();
 	}
 
+	/**
+	 * True while any utterance is still audible or synthesizing — a live
+	 * player, an unfinished stream handle, or an in-flight rewrite is enough.
+	 * Callers (Esc handler) treat this as the "silence me" signal.
+	 */
+	isSpeaking(): boolean {
+		return this.#livePlayers.size > 0 || this.#liveAborts.size > 0 || this.#handle !== null;
+	}
+
 	/** Lower the volume while the user is speaking (push-to-talk), so speech doesn't drown them out. */
 	duck(): void {
 		this.#ducked = true;

packages/coding-agent/test/input-controller-escape.test.ts

@@ -4,6 +4,7 @@ import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config
 import { InputController } from "@oh-my-pi/pi-coding-agent/modes/controllers/input-controller";
 import type { InteractiveModeContext, SubmittedUserInput } from "@oh-my-pi/pi-coding-agent/modes/types";
 import { USER_INTERRUPT_LABEL } from "@oh-my-pi/pi-coding-agent/session/messages";
+import { vocalizer } from "@oh-my-pi/pi-coding-agent/tts/vocalizer";
 import * as logger from "@oh-my-pi/pi-utils/logger";
 
 type Spy = Mock<(...args: unknown[]) => unknown>;
@@ -702,6 +703,27 @@ describe("InputController escape behavior", () => {
 		expect(ctx.showTreeSelector).not.toHaveBeenCalled();
 		expect(ctx.showUserMessageSelector).not.toHaveBeenCalled();
 	});
+
+	it("silences a still-audible vocalizer on Esc instead of opening the tree selector (#4521)", () => {
+		const clear = vi.spyOn(vocalizer, "clear").mockImplementation(() => {});
+		const isSpeaking = vi.spyOn(vocalizer, "isSpeaking").mockReturnValue(true);
+		const { ctx, editor, spies } = createContext();
+		const controller = new InputController(ctx);
+
+		controller.setupKeyHandlers();
+		editor.onEscape?.();
+
+		expect(clear).toHaveBeenCalledTimes(1);
+		expect(ctx.showTreeSelector).not.toHaveBeenCalled();
+		expect(ctx.showUserMessageSelector).not.toHaveBeenCalled();
+		expect(spies.resetDisplay).not.toHaveBeenCalled();
+
+		// A second Esc after silence must NOT immediately fire the double-Esc
+		// gesture — the first press consumed the arm.
+		isSpeaking.mockReturnValue(false);
+		editor.onEscape?.();
+		expect(ctx.showTreeSelector).not.toHaveBeenCalled();
+	});
 });
 
 describe("InputController Ctrl+C behavior", () => {