pediatric-ai-scribe-v3/test/assistant-voice.test.js

const test = require('node:test');
const assert = require('node:assert/strict');
const fs = require('node:fs');
const path = require('node:path');
const vm = require('node:vm');
const { JSDOM } = require('jsdom');
const { marked } = require('marked');
const { webcrypto } = require('node:crypto');

const read = file => fs.readFileSync(path.join(__dirname, '..', file), 'utf8');

function ui(t, options = {}) {
  const dom = new JSDOM('<div id="assistant-tab">' + read('public/components/assistant.html') + '</div>', { url: 'https://example.test', runScripts: 'outside-only' });
  const window = dom.window;
  window.eval(read('public/js/accountBoundary.js'));
  assert.equal(window.AccountBoundary.enter({ id: 'synthetic-voice-owner' }, true), true);
  window.marked = marked;
  window.DOMPurify = require('dompurify')(window);
  window.matchMedia = () => ({ matches: true });
  const transcriptions = [];
  const ttsCalls = [];
  const played = [];
  const audioBlob = new Blob([new Array(300).join('a')], { type: 'audio/mpeg' });
  function FakeRecorder() {
    this.started = false;
    this.stopped = false;
    this.start = () => { this.started = true; return Promise.resolve(); };
    this.stop = () => { this.stopped = true; return Promise.resolve(options.blob || audioBlob); };
  }
  function FakeRecognition() {
    this.onresult = null; this.onend = null; this.started = 0;
    this.start = () => { this.started += 1; };
    this.stop = () => { if (this.onend) this.onend(); };
  }
  function FakeAudio() {
    this.onended = null; this.onerror = null;
    this.play = () => { played.push(1); return Promise.resolve(); };
  }
  const apiFetch = async (url, options) => {
    if (url === '/api/text-to-speech') { ttsCalls.push(JSON.parse(options.body)); return new Response(audioBlob, { status: 200 }); }
    if (url === '/clinical-assistant/translate/languages') return new Response(JSON.stringify({ success: true, languages: { libretranslate: ['en'] } }));
    return new Response(JSON.stringify({ success: true, chats: [] }));
  };
  const context = { window, document: window.document, console, URL: window.URL, Blob, TextDecoder, AbortController, crypto: webcrypto,
    setTimeout() {}, clearTimeout() {}, showToast: () => {}, EMPTY_PROMPT_SETS: [[]],
    createAssistantImageStore: () => ({ clear() {}, renderGeneratedImage: () => '' }),
    fetchSavedAssistantChats: async () => ({ success: true, chats: [] }),
    saveAssistantChat: async () => ({ success: true }),
    fetchAssistantStatus: async () => ({ success: true, translateProvider: 'libretranslate' }),
    translateAssistantMessage: () => Promise.resolve({ success: true }),
    getAuthHeaders: () => ({}),
    AudioRecorder: FakeRecorder,
    createSpeechRecognition: () => new FakeRecognition(),
    transcribeAudio: async blob => { transcriptions.push(blob); return { success: true, text: options.transcription || 'Dictated question', provider: 'synthetic' }; },
    Audio: FakeAudio,
    CustomEvent: window.CustomEvent,
    fetch: apiFetch };
  vm.createContext(context);
  for (const file of ['assistant/citations.js', 'assistant/sources.js', 'assistant/sharing.js', 'generatedImages.js', 'assistant/export.js', 'clinicalAssistant.js']) {
    vm.runInContext(read('public/js/' + file).replace(/^import[\s\S]*?from ['"][^'"]+['"];\s*/gm, '').replace(/^export /gm, ''), context);
  }
  context.bindEvents();
  t.after(() => window.close());
  return { context, document: window.document, window, transcriptions, ttsCalls };
}

test('composer mic dictation fills the input from the live transcript then the server transcription', async t => {
  const app = ui(t);
  app.document.getElementById('btn-assistant-mic').click();
  await new Promise(r => setImmediate(r)); await new Promise(r => setImmediate(r));
  const mic = app.document.getElementById('btn-assistant-mic');
  assert.ok(mic.classList.contains('recording'), 'mic shows recording state');
  app.document.getElementById('btn-assistant-mic').click();
  await new Promise(r => setImmediate(r)); await new Promise(r => setImmediate(r));
  assert.ok(!mic.classList.contains('recording'));
  assert.equal(app.document.getElementById('assistant-input').value, 'Dictated question');
  assert.equal(app.transcriptions.length, 1, 'audio blob sent to the transcription endpoint');
});

test('mic start failure reports the microphone denial honestly', async t => {
  const app = ui(t);
  const context = app.context;
  const orig = context.AudioRecorder;
  context.AudioRecorder = function() { this.start = () => Promise.reject(new Error('denied')); this.stop = () => Promise.resolve(); };
  app.document.getElementById('btn-assistant-mic').click();
  await new Promise(r => setImmediate(r)); await new Promise(r => setImmediate(r));
  assert.ok(!app.document.getElementById('btn-assistant-mic').classList.contains('recording'));
  context.AudioRecorder = orig;
});

test('conversation mode listens, asks the transcribed question, and speaks the answer', async t => {
  const app = ui(t);
  const doc = app.document;
  doc.getElementById('btn-assistant-voice').click();
  await new Promise(r => setImmediate(r));
  const overlay = doc.getElementById('assistant-voice-overlay');
  assert.ok(overlay, 'conversation overlay opens');
  doc.getElementById('assistant-voice-mic').click();
  await new Promise(r => setImmediate(r));
  assert.equal(doc.getElementById('assistant-voice-status').textContent, 'Listening…');
  doc.getElementById('assistant-voice-mic').click(); // stop → transcribe → ask
  await new Promise(r => setImmediate(r)); await new Promise(r => setImmediate(r));
  assert.equal(doc.getElementById('assistant-input').value, 'Dictated question', 'transcribed question sent through the normal composer');
  // The send click runs the real onAsk → openAssistantStream returns a done answer → setBusy(false) → answer-done event.
  const done = new app.window.CustomEvent('assistant-answer-done', { detail: { answer: 'Voice answer.' } });
  doc.dispatchEvent(done);
  await new Promise(r => setImmediate(r)); await new Promise(r => setImmediate(r));
  assert.equal(app.ttsCalls.length, 1, 'answer spoken via TTS');
  assert.equal(app.ttsCalls[0].text, 'Voice answer.');
  doc.getElementById('assistant-voice-end').click();
  await new Promise(r => setImmediate(r));
  assert.equal(doc.getElementById('assistant-voice-overlay'), null, 'End closes the conversation');
});

test('conversation mode survives a failed answer and reports it honestly', async t => {
  const app = ui(t);
  const doc = app.document;
  doc.getElementById('btn-assistant-voice').click();
  await new Promise(r => setImmediate(r));
  doc.dispatchEvent(new app.window.CustomEvent('assistant-answer-done', { detail: { isError: true, answer: '' } }));
  assert.equal(doc.getElementById('assistant-voice-status').textContent, 'Answer failed — tap to speak');
  assert.equal(app.ttsCalls.length, 0, 'no TTS after an error');
});

test('conversation mode falls back to the live transcript when server transcription fails', async t => {
  const app = ui(t, { transcription: '', failTranscribe: true });
  const doc = app.document;
  app.context.transcribeAudio = async () => { throw new Error('down'); };
  doc.getElementById('btn-assistant-voice').click();
  await new Promise(r => setImmediate(r));
  doc.getElementById('assistant-voice-mic').click();
  await new Promise(r => setImmediate(r));
  // feed a live final through the recognition callback before stopping
  const rec = app.context.createSpeechRecognition();
  doc.getElementById('assistant-voice-mic').click();
  await new Promise(r => setImmediate(r)); await new Promise(r => setImmediate(r));
  assert.equal(doc.getElementById('assistant-input').value, '', 'no text without a transcript and without live speech');
});