import { Container, stripTerminalSequences, type TUI, type TuiMouseEvent, } from "@earendil-works/pi-tui"; import { afterEach, describe, expect, it, vi } from "vitest"; import { createNativeVoiceWidget, type NativeVoiceWidgetState, renderNativeVoiceWidget, widgetVisibleWidth, } from "../src/voice-widget.ts"; const theme = { fg: (_color: string, text: string) => `\u001b[31m${text}\u001b[0m`, bold: (text: string) => text, } as never; function state( overrides: Partial = {}, ): NativeVoiceWidgetState { return { status: "ready", activity: "ready", error: null, partial: "", spoken: "", waveform: [0, 0.5, 1], inputLevel: 0, outputLevel: 0, playing: false, muted: false, ...overrides, }; } function fixture(input = state()) { const requestRender = vi.fn(); const terminal = { rows: 24 }; const widget = createNativeVoiceWidget( { requestRender, terminal } as unknown as TUI, theme, () => input, ); return { input, widget, requestRender, terminal, text: (width = 60) => widget.render(width).map(stripTerminalSequences), }; } function wheel(y: number, wheelDelta: number, height = 7): TuiMouseEvent { return { type: "wheel", button: "none", x: 5, y, screenX: 5, screenY: y, width: 60, height, shift: false, alt: false, ctrl: false, wheelDelta, }; } const numbered = (prefix: string, count: number) => Array.from({ length: count }, (_, i) => `${prefix}${i}`).join("\n"); afterEach(() => vi.useRealTimers()); describe("native voice widget", () => { it("keeps non-speech flat and muted despite changing microphone peaks", () => { const f = fixture( state({ humanSpeaking: false, waveform: [0.1, 0.8, 0.01] }), ); const idle = f.text()[0]; f.input.waveform = [0.9, 0.01, 0.4]; f.input.inputLevel = 0.9; f.input.outputLevel = 0.8; expect(f.text()[0]).toBe(idle); expect(idle).toContain("───"); expect(idle).not.toMatch(/[▁▂▃▄▅▆▇█]/u); f.widget.dispose(); }); it("amplifies detected speech visually without changing measured audio", () => { const input = state({ humanSpeaking: true, waveform: [0.001, 0.01, 0.1] }); const bold = vi.fn((text: string) => text); const fg = vi.fn((_color: string, text: string) => text); expect(renderNativeVoiceWidget(input, { bold, fg }, 80)[0]).toContain( "▂▅█", ); expect(bold).toHaveBeenCalledWith("▂▅█"); expect(fg).toHaveBeenCalledWith("accent", "▂▅█"); expect(input.waveform).toEqual([0.001, 0.01, 0.1]); input.muted = true; expect(renderNativeVoiceWidget(input, { bold, fg }, 80)[0]).toContain( "───", ); expect(fg).toHaveBeenCalledWith("muted", "───"); }); it("keeps startup compact and expands only when the call is ready", () => { const f = fixture(state({ status: "loading" })); expect(f.text()).toHaveLength(2); expect(f.text()[0]).toContain("Loading"); expect(f.text()[1]).toContain("Starting speech"); expect(f.text()[1]).toContain("/voice stop"); expect(f.text().join("\n")).not.toMatch( /Paused|Nothing spoken|Provisioning/, ); expect(f.widget.handleMouse?.(wheel(1, 1))).toBeUndefined(); f.input.status = "ready"; expect(f.text()).toHaveLength(7); f.widget.dispose(); }); it("lowercases provisional display only, preserving raw text, final casing and spoken captions", () => { const input = state({ partial: "USE ONNX CAFÉ İ 你好", spoken: "Keep ONNX casing.", provisional: true, }); const f = fixture(input); for (const lines of [ f.text(), renderNativeVoiceWidget(input, theme, 60).map(stripTerminalSequences), ]) { expect(lines.slice(1, 4).join(" ")).toContain( "use onnx café i\u0307 你好", ); expect(lines.slice(4).join(" ")).toContain("Keep ONNX casing."); } expect(input.partial).toBe("USE ONNX CAFÉ İ 你好"); input.provisional = false; expect(f.text()[1]).toContain("USE ONNX CAFÉ İ 你好"); for (const width of [1, 8, 20]) for (const line of f.text(width)) expect(widgetVisibleWidth(line)).toBeLessThanOrEqual(width); f.widget.dispose(); }); it("uses Nerd Font icons without jargon or a command footer", () => { const lines = renderNativeVoiceWidget( state({ partial: "hello", spoken: "welcome" }), theme, 60, ).map(stripTerminalSequences); expect(lines).toHaveLength(7); expect(lines[1]).toContain(" hello"); expect(lines[4]).toContain(" welcome"); expect(lines.join("\n")).not.toMatch(/ASR|SAY|\/voice help/); expect( renderNativeVoiceWidget(state({ muted: true }), theme, 60).join("\n"), ).toContain(""); }); it.each([ { activity: "ready", playing: false, label: " Listening" }, { activity: "ready", playing: true, label: " Speaking" }, { activity: "thinking", playing: false, label: "󰧑 Thinking" }, { activity: "thinking", playing: true, label: " Speaking · 󰧑 Thinking" }, ])("shows Nerd Font state icons: $label", ({ activity, playing, label }) => { const input = state({ activity, playing }); const f = fixture(input); for (const first of [ f.text(80)[0], ...renderNativeVoiceWidget(input, theme, 80) .map(stripTerminalSequences) .slice(0, 1), ]) expect(first).toContain(label); expect(f.text().join("\n")).not.toMatch(/[🎤🔇🔊💭]/u); f.widget.dispose(); }); it("keeps ANSI, emoji, combining and wide Unicode output within resized widths", () => { const input = state({ partial: "你好 👨‍👩‍👧‍👦 café e\u0301 world", spoken: "Dziękuję. こんにちは。", }); for (const width of [0, 1, 2, 8, 20, 80, Number.NaN, Infinity]) for (const line of renderNativeVoiceWidget(input, theme, width)) expect(widgetVisibleWidth(line)).toBeLessThanOrEqual(width); }); it("wraps rather than truncating and exposes text beyond the old 256-character cutoff", () => { const f = fixture( state({ partial: `BEGIN ${"word ".repeat(100)}END_TOKEN` }), ); expect(f.text(20).slice(1, 4).join(" ")).toContain("END_TOKEN"); f.widget.scroll("heard", -10000); expect(f.text(20)[1]).toContain("BEGIN"); expect( f .text(20) .slice(1, 4) .every((line) => !line.includes("…")), ).toBe(true); f.widget.dispose(); }); it("scrolls through the mounted Container path without requesting editor focus", () => { const f = fixture( state({ partial: numbered("heard", 10), spoken: numbered("spoken", 10) }), ); const mounted = new Container(); mounted.addChild({ render: () => ["other widget"], invalidate() {} }); mounted.addChild(f.widget); mounted.render(60); const result = mounted.handleMouse(wheel(5, -2, 8)); expect(result?.handled).toBe(true); expect(result?.focus).not.toBe(true); expect(f.widget.handleInput).toBeUndefined(); const lines = f.text(); expect(lines[1]).toContain("heard7"); expect(lines[4]).toContain("spoken5"); const boundary = mounted.handleMouse(wheel(5, 1000, 8)); expect(boundary?.handled).toBe(true); expect(f.widget.handleMouse?.(wheel(0, 1))).toBeUndefined(); f.widget.dispose(); }); it("holds the scrolled region while new text arrives and resumes follow at the bottom", () => { const f = fixture( state({ partial: numbered("heard", 10), spoken: numbered("spoken", 10) }), ); f.text(); f.widget.handleMouse?.(wheel(2, -3)); f.input.partial += "\nheard10"; f.input.spoken += "\nspoken10"; let lines = f.text(); expect(lines[1]).toContain("heard4"); expect(lines[4]).toContain("spoken8"); expect(lines[1]).toContain("↑"); expect(lines[3]).toContain("↓"); f.widget.scroll("heard", 10000); f.input.partial += "\nheard11"; lines = f.text(); expect(lines[1]).toContain("heard9"); expect(lines[3]).toContain("heard11"); f.widget.dispose(); }); it("adapts height to short terminals and clamps scrolling after replacement and resize", () => { const f = fixture(state({ spoken: numbered("caption", 30) })); f.text(); f.widget.scroll("spoken", -10000); f.terminal.rows = 12; expect(f.text(20)).toHaveLength(4); f.input.spoken = "tiny"; // 10 columns leave 8 for content inside the one-column padding. expect(f.text(10).join("\n")).toContain("tiny"); for (const line of f.text(10)) { expect(line.startsWith(" ")).toBe(true); expect(widgetVisibleWidth(line)).toBeLessThanOrEqual(10); } f.widget.dispose(); }); it("bounds caption layout work and strips terminal controls while retaining newlines", () => { const f = fixture( state({ partial: `\u001b]0;untrusted title\u0007\u001b[31mfirst\u001b[0m\nsecond\u0000`, spoken: "x".repeat(50000), }), ); const lines = f.text(); expect(lines[1]).toContain("first"); expect(lines[2]).toContain("second"); expect(lines.join("\n")).not.toContain("\u0000"); expect(lines.join("\n")).not.toContain("untrusted title"); expect(lines[6]).toContain("…"); f.widget.dispose(); }); it("throttles render requests and cancels pending work on disposal", () => { vi.useFakeTimers(); const f = fixture(); f.widget.invalidate(); f.widget.invalidate(); f.widget.invalidate(); expect(f.requestRender).toHaveBeenCalledOnce(); vi.advanceTimersByTime(100); expect(f.requestRender).toHaveBeenCalledTimes(2); f.widget.invalidate(); f.widget.dispose(); vi.advanceTimersByTime(100); expect(f.requestRender).toHaveBeenCalledTimes(2); expect(f.widget.handleMouse?.(wheel(2, 1))).toBeUndefined(); }); it("dims provisional captions and reveals final text before bounded accent sweep completes", () => { vi.useFakeTimers(); const colorTheme = { fg: (color: string, text: string) => `\u001b[${color === "muted" ? "32" : "31"}m${text}\u001b[0m`, bold: (text: string) => text, } as never; const f = fixture( state({ partial: "draft", provisional: true, captionRevision: 0 }), ); const widget = createNativeVoiceWidget( { requestRender: f.requestRender, terminal: f.terminal, } as unknown as TUI, colorTheme, () => f.input, ); const draft = widget.render(60)[1]; expect(draft).toContain("[32m"); f.input.partial = "final café 👩‍💻 你好"; f.input.provisional = false; f.input.captionRevision = 1; const final = widget.render(60)[1] ?? ""; expect(stripTerminalSequences(final)).toContain("final café 👩‍💻 你好"); expect(final).toContain("[31m"); for (let frame = 0; frame < 5; frame++) { vi.advanceTimersByTime(50); expect(stripTerminalSequences(widget.render(60)[1] ?? "")).toContain( "final café 👩‍💻 你好", ); } vi.advanceTimersByTime(50); widget.render(60); expect(vi.getTimerCount()).toBe(0); widget.dispose(); f.widget.dispose(); }); it("cancels accent sweep on new provisional speech and disposal without changing scroll", () => { vi.useFakeTimers(); const f = fixture( state({ partial: numbered("heard", 10), provisional: true, captionRevision: 0, }), ); f.text(); f.widget.scroll("heard", -3); const before = f.text()[1]; f.input.partial = numbered("final", 10); f.input.provisional = false; f.input.captionRevision = 1; f.widget.render(60); f.input.partial = numbered("later", 10); f.input.provisional = true; f.widget.render(60); vi.advanceTimersByTime(300); expect((f.text()[1] ?? "").replace(/later(\d+)/u, "heard$1")).toBe(before); f.input.provisional = false; f.input.captionRevision = 2; f.widget.render(60); expect(vi.getTimerCount()).toBe(1); f.widget.dispose(); expect(vi.getTimerCount()).toBe(0); }); it("shows actual unavailable and reconnecting state", () => { expect( renderNativeVoiceWidget( state({ status: "reconnecting" }), theme, 80, ).join("\n"), ).toContain("Reconnecting"); expect( renderNativeVoiceWidget( state({ activity: "thinking", error: "device lost" }), theme, 80, ).join("\n"), ).toContain("Error: device lost"); }); });