Files
learnharness/tests/dictation-flow.test.tsx

391 lines
18 KiB
TypeScript

import React, { act } from "react";
import assert from "node:assert/strict";
import { afterEach, test } from "node:test";
import { parseHTML } from "linkedom";
import { finishDictation, useDictation, type Dictation } from "../components/dictation.tsx";
import { I18nProvider, useI18n } from "../components/i18n.tsx";
import { notifySettingsSaved } from "../components/settings-signal.ts";
import { isDictationActive, useDictationActive } from "../components/dictation-state.ts";
import type { GeneralSettings, UserSettings } from "../src/core/index.ts";
function deferred<T>() {
let resolve!: (value: T) => void;
const promise = new Promise<T>((done) => { resolve = done; });
return { promise, resolve };
}
class Socket extends EventTarget {
static CONNECTING = 0;
static OPEN = 1;
static all: Socket[] = [];
readyState = 0;
sent: unknown[] = [];
constructor(public url: string) { super(); Socket.all.push(this); }
get language() { return new URL(this.url).searchParams.get("language"); }
open() { this.readyState = 1; this.dispatchEvent(new Event("open")); }
send(value: unknown) { this.sent.push(value); }
close() { this.readyState = 3; }
end(code = 1000) { this.close(); this.dispatchEvent(Object.assign(new Event("close"), { code })); }
final(text: string) {
this.dispatchEvent(Object.assign(new Event("message"), {
data: JSON.stringify({ type: "Results", is_final: true, channel: { alternatives: [{ transcript: text }] } }),
}));
}
}
class AudioGraph {
state = "running";
destination = {};
audioWorklet = { addModule: async () => {} };
createMediaStreamSource() { return { connect() {}, disconnect() {} }; }
}
class Worklet {
static latest: Worklet;
static autoFlush = true;
disconnected = false;
port = {
onmessage: (_event: { data: unknown }) => {},
postMessage: (_message: unknown) => { if (Worklet.autoFlush) queueMicrotask(() => this.flush()); },
};
constructor() { Worklet.latest = this; }
flush() {
this.port.onmessage({ data: { type: "audio", buffer: new Int16Array([7]).buffer } });
this.port.onmessage({ data: { type: "flushed" } });
}
connect() {}
disconnect() { this.disconnected = true; }
}
const globals = globalThis as Record<string, unknown>;
let restore: (() => void) | undefined;
let dispose: (() => Promise<void>) | undefined;
async function fixture(general: GeneralSettings = { sttLanguage: "ru" }) {
const dom = parseHTML("<html><body><div id='root'></div></body></html>");
const state = {
// The contradictory upstream preference must never control recording.
record: { version: 1, general, dictation: { language: "pl" } } as UserSettings,
nextRead: undefined as Promise<Response> | undefined,
nextWrite: undefined as Promise<Response> | undefined,
nextMic: undefined as Promise<{ getTracks: () => { stop: () => void }[] }> | undefined,
nextToken: undefined as (() => Promise<Response>) | undefined,
readFails: false,
reads: 0,
active: false,
captures: 0,
stops: 0,
appends: [] as string[],
};
Socket.all = [];
Worklet.autoFlush = true;
const previous = new Map<string, PropertyDescriptor | undefined>();
for (const [key, value] of Object.entries({
React, window: dom.window, document: dom.document, CustomEvent: dom.window.CustomEvent,
navigator: { userAgent: "node", languages: ["en-US"], mediaDevices: { getUserMedia: async () => {
state.captures++;
if (state.nextMic) return state.nextMic;
return { getTracks: () => [{ stop() { state.stops++; } }] };
} } },
WebSocket: Socket, AudioContext: AudioGraph, AudioWorkletNode: Worklet,
IS_REACT_ACT_ENVIRONMENT: true,
fetch: async (url: string, init?: RequestInit) => {
if (url === "/api/stt/token") return state.nextToken ? state.nextToken() : Response.json({ token: "test-token" });
if (url === "/api/settings") {
state.reads++;
if (state.nextRead) { const response = state.nextRead; state.nextRead = undefined; return response; }
return state.readFails ? Response.json({}, { status: 500 }) : Response.json({ settings: state.record });
}
assert.equal(url, "/api/settings/general");
assert.equal(init?.method, "PUT");
if (state.nextWrite) { const response = state.nextWrite; state.nextWrite = undefined; return response; }
state.record = { ...state.record, general: JSON.parse(init!.body as string) };
return Response.json({ settings: state.record });
},
})) {
previous.set(key, Object.getOwnPropertyDescriptor(globalThis, key));
Object.defineProperty(globalThis, key, { configurable: true, writable: true, value });
}
restore = () => {
for (const [key, descriptor] of previous) {
if (descriptor) Object.defineProperty(globalThis, key, descriptor);
else delete globals[key];
}
};
const { createRoot } = await import("react-dom/client");
let latest!: Dictation;
let setting!: ReturnType<typeof useI18n>;
function Harness() {
latest = useDictation({ append: (text) => state.appends.push(text) });
setting = useI18n();
state.active = useDictationActive();
return null;
}
let root = createRoot(document.getElementById("root")!);
const mount = async () => {
await act(async () => { root.render(<I18nProvider initialGeneral={state.record.general}><Harness /></I18nProvider>); });
};
dispose = async () => { await act(async () => root.unmount()); };
return {
state,
get dictation() { return latest; },
get setting() { return setting; },
mount,
unmount: () => dispose!(),
remount: async () => { await dispose!(); root = createRoot(document.getElementById("root")!); await mount(); },
};
}
afterEach(async () => { await dispose?.(); restore?.(); dispose = undefined; restore = undefined; });
test("general language and hints are snapshotted; reconnect and final delivery retain them", async () => {
const f = await fixture({ sttLanguage: "multi", sttPreferredLanguages: ["ru", "en"] });
await f.mount();
await act(async () => { f.dictation.toggle(); });
assert.equal(f.state.active, true);
const first = Socket.all[0]!;
assert.equal(first.language, "multi");
assert.deepEqual(new URL(first.url).searchParams.getAll("language_hint"), ["ru", "en"]);
await act(async () => { first.open(); first.final("Первое"); });
await act(async () => { await f.setting.saveGeneral({ sttLanguage: "en", sttPreferredLanguages: [] }); });
await act(async () => { first.end(1006); });
const retry = Socket.all[1]!;
assert.equal(retry.url, first.url);
await act(async () => { retry.open(); retry.final("second"); });
let finished = false;
let done!: Promise<void>;
await act(async () => { done = f.dictation.finish().then(() => { finished = true; }); });
assert.equal(finished, false);
assert.deepEqual(f.state.appends, []);
assert.equal(retry.sent.at(-1), '{"type":"CloseStream"}');
assert(retry.sent.at(-2) instanceof ArrayBuffer, "the final PCM frame precedes CloseStream");
await act(async () => { retry.final("last"); retry.end(); await done; });
assert.deepEqual(f.state.appends, ["Первое second last"]);
assert.equal(f.state.active, false);
await act(async () => { retry.final("stale"); retry.end(); });
assert.deepEqual(f.state.appends, ["Первое second last"]);
await act(async () => { f.dictation.toggle(); });
assert.equal(Socket.all[2]!.language, "en");
});
for (const [selection, expected] of [[undefined, "en"], ["en", "en"], ["ru", "ru"], ["multi", "multi"]] as const) {
test(`fork language ${selection ?? "system"} ignores contradictory dictation.language`, async () => {
const f = await fixture({ sttLanguage: selection });
await f.mount();
await act(async () => { f.dictation.toggle(); });
assert.equal(Socket.all[0]!.language, expected);
assert.equal(f.state.reads, 0, "recording does not load an independent preference");
});
}
test("a failed general save retains the confirmed language, including a shortcut during saving", async () => {
const f = await fixture(); await f.mount();
const write = deferred<Response>(); f.state.nextWrite = write.promise;
let saved!: Promise<void>;
await act(async () => { saved = f.setting.saveGeneral({ sttLanguage: "multi" }); saved.catch(() => {}); });
await act(async () => { window.dispatchEvent(new window.Event("lh:kbd-toggle-dictation")); });
assert.equal(Socket.all[0]!.language, "ru");
await act(async () => { write.resolve(Response.json({}, { status: 500 })); await assert.rejects(saved, /Could not save/); });
assert.equal(f.setting.sttLanguage, "ru");
});
test("confirmed general choices survive navigation and a failed refresh does not substitute upstream defaults", async () => {
const f = await fixture(); await f.mount();
await act(async () => { await f.setting.saveGeneral({ sttLanguage: "multi" }); });
await f.remount();
f.state.readFails = true;
await act(async () => { notifySettingsSaved("general"); });
await act(async () => { f.dictation.toggle(); });
assert.equal(Socket.all[0]!.language, "multi");
assert.equal(f.dictation.error, null);
});
test("capture uses the confirmed general choice while a save is pending and the new one on the next run", async () => {
const f = await fixture(); await f.mount();
const write = deferred<Response>(); f.state.nextWrite = write.promise;
let saved!: Promise<void>;
await act(async () => { saved = f.setting.saveGeneral({ sttLanguage: "multi" }); });
await act(async () => { f.dictation.toggle(); });
assert.equal(Socket.all[0]!.language, "ru");
f.state.record.general = { sttLanguage: "multi" };
await act(async () => { write.resolve(Response.json({ settings: f.state.record })); await saved; });
await act(async () => { f.dictation.cancel(); f.dictation.toggle(); });
assert.equal(Socket.all[1]!.language, "multi");
});
test("a stale general refresh cannot overwrite a newer confirmed choice", async () => {
const f = await fixture(); await f.mount();
const read = deferred<Response>(); f.state.nextRead = read.promise;
await act(async () => { notifySettingsSaved("general"); });
await act(async () => { await f.setting.saveGeneral({ sttLanguage: "multi" }); });
await act(async () => { read.resolve(Response.json({ settings: { general: { sttLanguage: "ru" } } })); });
assert.equal(f.setting.sttLanguage, "multi");
await act(async () => { f.dictation.toggle(); });
assert.equal(Socket.all[0]!.language, "multi");
});
test("cancel while microphone permission is pending prevents orphaned capture", async () => {
const f = await fixture(); await f.mount();
const mic = deferred<{ getTracks: () => { stop: () => void }[] }>(); f.state.nextMic = mic.promise;
await act(async () => { f.dictation.toggle(); f.dictation.cancel(); });
await act(async () => { mic.resolve({ getTracks: () => [{ stop() { f.state.stops++; } }] }); });
assert.equal(f.state.stops, 1);
assert.equal(Socket.all.length, 0);
assert.equal(isDictationActive(), false);
});
test("finish in the start turn cancels a pending microphone without hanging", async () => {
const f = await fixture(); await f.mount();
const mic = deferred<{ getTracks: () => { stop: () => void }[] }>(); f.state.nextMic = mic.promise;
await act(async () => { f.dictation.toggle(); await finishDictation(); });
assert.equal(f.dictation.phase, "idle");
await act(async () => { mic.resolve({ getTracks: () => [{ stop() { f.state.stops++; } }] }); });
assert.equal(f.state.stops, 1);
assert.equal(Socket.all.length, 0);
});
test("Stop while the socket connects drains PCM and waits for the final transcript", async () => {
const f = await fixture(); await f.mount();
await act(async () => { f.dictation.toggle(); });
const socket = Socket.all[0]!;
let done!: Promise<void>; let finished = false;
await act(async () => { done = finishDictation().then(() => { finished = true; }); });
assert.equal(f.dictation.phase, "stopping");
assert.equal(finished, false);
await act(async () => { socket.open(); });
assert.equal(f.dictation.phase, "stopping");
assert(socket.sent[0] instanceof ArrayBuffer);
assert.equal(socket.sent[1], '{"type":"CloseStream"}');
await act(async () => { socket.final("tail"); socket.end(); await done; });
assert.deepEqual(f.state.appends, ["tail"]);
});
test("a socket opening during drain cannot close the stream before the tail PCM", async () => {
const f = await fixture(); await f.mount(); Worklet.autoFlush = false;
await act(async () => { f.dictation.toggle(); });
const socket = Socket.all[0]!;
let done!: Promise<void>;
await act(async () => { done = finishDictation(); socket.open(); });
assert.equal(socket.sent.length, 0);
assert.equal(f.dictation.phase, "stopping");
await act(async () => { Worklet.latest.flush(); });
assert(socket.sent[0] instanceof ArrayBuffer);
assert.equal(socket.sent[1], '{"type":"CloseStream"}');
await act(async () => { socket.final("complete"); socket.end(); await done; });
assert.deepEqual(f.state.appends, ["complete"]);
});
test("a reconnect while stopping retains its session and still finishes exactly once", async () => {
const f = await fixture(); await f.mount();
await act(async () => { f.dictation.toggle(); });
const socket = Socket.all[0]!;
await act(async () => { socket.open(); socket.final("before"); });
let done!: Promise<void>;
await act(async () => { done = finishDictation(); });
await act(async () => { socket.end(1006); });
const retry = Socket.all[1]!;
assert.equal(retry.url, socket.url);
await act(async () => { retry.open(); });
assert.equal(retry.sent.at(-1), '{"type":"CloseStream"}');
assert.equal(f.dictation.phase, "stopping");
await act(async () => { retry.final("after"); retry.end(); await done; });
assert.deepEqual(f.state.appends, ["before after"]);
assert.equal(f.state.active, false);
});
test("cancel during drain resolves finish and ignores late worklet/socket messages", async () => {
const f = await fixture(); await f.mount(); Worklet.autoFlush = false;
await act(async () => { f.dictation.toggle(); });
await act(async () => { Socket.all[0]!.open(); Socket.all[0]!.final("discard"); });
let done!: Promise<void>;
await act(async () => { done = f.dictation.finish(); f.dictation.cancel(); await done; });
await act(async () => { Worklet.latest.flush(); Socket.all[0]!.final("late"); Socket.all[0]!.end(1006); });
assert.deepEqual(f.state.appends, []);
assert.equal(Socket.all.length, 1);
assert.equal(f.state.active, false);
});
test("unmount during a pending general refresh prevents orphaned capture", async () => {
const f = await fixture(); await f.mount();
const read = deferred<Response>(); f.state.nextRead = read.promise;
await act(async () => { notifySettingsSaved("general"); });
await f.unmount();
await act(async () => { read.resolve(Response.json({ settings: f.state.record })); });
assert.equal(f.state.captures, 0);
assert.equal(Socket.all.length, 0);
assert.equal(isDictationActive(), false);
});
test("unmount releases an active stream/socket and discards deferred finals", async () => {
const f = await fixture(); await f.mount();
await act(async () => { f.dictation.toggle(); });
const socket = Socket.all[0]!;
await act(async () => { socket.open(); socket.final("do not append into another page"); });
await f.unmount();
assert.equal(f.state.stops, 1);
assert.equal(socket.readyState, 3);
assert.equal(isDictationActive(), false);
assert.deepEqual(f.state.appends, []);
await act(async () => { socket.final("late event"); socket.end(1006); });
assert.equal(Socket.all.length, 1);
assert.deepEqual(f.state.appends, []);
});
test("system STT follows the fork's detected Russian locale, not dictation.language", async () => {
const f = await fixture({});
Object.defineProperty(globalThis, "navigator", { configurable: true, value: {
userAgent: "node", languages: ["ru-RU"],
mediaDevices: { getUserMedia: async () => ({ getTracks: () => [{ stop() { f.state.stops++; } }] }) },
} });
await f.mount();
await act(async () => { f.dictation.toggle(); });
assert.equal(Socket.all[0]!.language, "ru");
});
test("a failed token request releases capture and leaves an honest error", async () => {
const f = await fixture();
f.state.nextToken = async () => Response.json({ message: "token denied" }, { status: 500 });
await f.mount();
await act(async () => { f.dictation.toggle(); });
assert.equal(f.dictation.error, "token denied");
assert.equal(f.dictation.phase, "idle");
assert.equal(f.state.stops, 1);
assert.equal(Socket.all.length, 0);
assert.equal(isDictationActive(), false);
});
test("one reconnect is allowed and a second socket failure ends the run", async () => {
const f = await fixture(); await f.mount();
await act(async () => { f.dictation.toggle(); });
await act(async () => { Socket.all[0]!.open(); Socket.all[0]!.final("kept"); Socket.all[0]!.end(1006); });
assert.equal(Socket.all.length, 2);
await act(async () => { Socket.all[1]!.end(1006); });
assert.equal(f.dictation.error, "Deepgram connection lost (1006)");
assert.equal(f.dictation.phase, "idle");
assert.deepEqual(f.state.appends, ["kept"]);
assert.equal(f.state.stops, 1);
});
test("a missing drain acknowledgement fails visibly and resolves submit waiters", async () => {
const f = await fixture(); await f.mount(); Worklet.autoFlush = false;
await act(async () => { f.dictation.toggle(); });
await act(async () => { Socket.all[0]!.open(); });
let done!: Promise<void>;
await act(async () => { done = finishDictation(); });
await act(async () => { await done; });
assert.equal(f.dictation.error, "Could not finish microphone capture");
assert.equal(f.dictation.phase, "idle");
assert.equal(Worklet.latest.disconnected, true);
assert.equal(isDictationActive(), false);
});
test("microphone permission resolving after unmount immediately stops the late stream", async () => {
const f = await fixture(); await f.mount();
const mic = deferred<{ getTracks: () => { stop: () => void }[] }>(); f.state.nextMic = mic.promise;
await act(async () => { f.dictation.toggle(); });
await f.unmount();
await act(async () => { mic.resolve({ getTracks: () => [{ stop() { f.state.stops++; } }] }); });
assert.equal(f.state.stops, 1);
assert.equal(Socket.all.length, 0);
assert.equal(isDictationActive(), false);
});