hub/tools/transcript-test.js

// Transcripts: reading SRT/VTT/JSON, finding an episode's transcript link in a feed, refusing private addresses, and speech-to-text through a fake Whisper server.
const http = require('http'), fs = require('fs'), os = require('os'), path = require('path');
const tr = require('../transcripts');
let bad = 0; const ok = (c, m) => { console.log((c ? 'PASS  ' : 'FAIL  ') + m); if (!c) bad++; };

(async () => {
  const dir = fs.mkdtempSync(path.join(os.tmpdir(), 'trt-'));
  let installed = false; const calls = [];
  const srv = http.createServer((req, res) => {
    const body = []; req.on('data', d => body.push(d)); req.on('end', () => {
      const b = Buffer.concat(body).toString('latin1');
      calls.push(req.method + ' ' + req.url);
      if (req.method === 'GET' && req.url === '/v1/models/Systran/faster-whisper-small') { res.writeHead(installed ? 200 : 404); return res.end('{}'); }
      if (req.method === 'POST' && req.url === '/v1/models') { installed = /faster-whisper-small/.test(b); res.writeHead(installed ? 201 : 422); return res.end('{}'); } // (this version of the service downloads through a body, the older path form is refused)
      if (req.url === '/v1/audio/transcriptions' && req.method === 'POST') {
        if (!installed) { res.writeHead(404); return res.end('{"detail":"Model is not installed locally"}'); }
        if (!/response_format/.test(b) || !/faster-whisper-small/.test(b) || !/hello-audio/.test(b)) { res.writeHead(400); return res.end('bad form'); }
        res.writeHead(200, { 'content-type': 'application/json' }); return res.end(JSON.stringify({ text: 'x', segments: [{ start: 0, end: 2.5, text: ' Welcome to the show. ' }, { start: 2.5, end: 6, text: ' Today we talk about <b>bread</b>.' }, { start: 6, end: 7, text: '  ' }] }));
      }
      res.writeHead(404); res.end();
    });
  });
  await new Promise(r => srv.listen(0, '127.0.0.1', r));
  tr.init({ dataDir: dir, apiUrl: 'http://127.0.0.1:' + srv.address().port, model: 'Systran/faster-whisper-small', port: 8080 });

  const srt = '1\n00:00:01,500 --> 00:00:03,000\nHello <i>there</i>\n\n2\n00:01:02,000 --> 00:01:05,250\nSecond &amp; last\n';
  let l = tr.parseCues(srt); ok(l.length === 2 && l[0].t === 1500 && l[0].text === 'Hello there' && l[1].t === 62000 && l[1].text === 'Second & last', 'SubRip cues are read, tags and entities removed');
  l = tr.parseCues('WEBVTT\n\n00:05.000 --> 00:07.000\nShort form\n\n1:00:00.000 --> 1:00:02.000 align:start\nAn hour in\n'); ok(l.length === 2 && l[0].t === 5000 && l[1].t === 3600000, 'WebVTT cues, short and hour times');
  l = tr.parseJson(JSON.stringify({ version: '1.0.0', segments: [{ startTime: 0, endTime: 4, speaker: 'Ann', body: 'Hi all' }, { startTime: 4.5, body: 'Next' }] })); ok(l.length === 2 && l[0].text === 'Ann: Hi all' && l[1].t === 4500, 'Podcast 2.0 JSON segments');
  ok(tr.parseJson('not json').length === 0, 'broken JSON gives nothing, not a crash');
  const shaped = tr.shape([{ t: 3000, text: 'b' }, { t: 1000, text: 'a' }], 'S'); ok(shaped.lines[0].text === 'a' && shaped.lines[0].e === 3000 && shaped.sync === 'line', 'lines are sorted and given ends');

  const feed = `<item><guid>g1</guid><podcast:transcript url="https://a/b.srt?x=1&amp;y=2" type="application/x-subrip"/></item><item><guid>g2</guid><enclosure url="https://f/2.mp3" length="1"/><podcast:transcript url="https://a/c.vtt" type="text/vtt"/></item>`;
  ok(tr.transcriptLinks(feed, 'g1', '')[0].url === 'https://a/b.srt?x=1&y=2', 'a transcript link is found by the episode guid (and & unescaped)');
  ok(tr.transcriptLinks(feed, 'nope', 'https://f/2.mp3')[0].type === 'text/vtt', 'or by the audio file address');
  ok(tr.transcriptLinks(feed, 'zzz', '') === null, 'an episode that is not in the feed is told apart from one with no transcript');
  ok(tr.transcriptLinks('<item><guid>q</guid></item>', 'q', '').length === 0, 'an episode with no transcript tag gives an empty list');

  for (const u of ['http://127.0.0.1:1/x', 'http://localhost/x', 'http://192.168.1.5/x', 'http://10.0.0.2/x', 'ftp://example.com/x', 'http://[::1]/x']) {
    let e = null; try { await tr.getText(u); } catch (x) { e = x; } ok(e && /not allowed/.test(e.message), 'refuses ' + u);
  }

  const audio = path.join(dir, 'a.bin'); fs.writeFileSync(audio, 'hello-audio');
  const r = await tr.whisper(audio, 'audio'); ok(r && r.found && r.lines.length === 2 && r.lines[0].text === 'Welcome to the show.' && r.lines[1].text === 'Today we talk about bread.' && r.source === 'Transcribed on your server', 'speech-to-text segments become timed lines (blank ones dropped, tags removed)');
  ok(calls.includes('POST /v1/models/Systran/faster-whisper-small') && calls.includes('POST /v1/models') && installed, 'a model that is not installed is downloaded first (trying the older form, then the newer)');
  ok(r.lines[1].t === 2500 && r.lines[1].e === 2500 + 6000, 'times are in milliseconds');
  srv.close(); fs.rmSync(dir, { recursive: true, force: true });
  console.log(bad ? '\n' + bad + ' check(s) failed' : '\nAll transcript checks passed'); process.exit(bad ? 1 : 0);
})().catch(e => { console.error(e); process.exit(1); });