nio-cat 0.9.5 → 0.9.7

package/src/voice.js

@@ -2,7 +2,9 @@
 // Veu: transcripció (STT) i lectura en veu alta (TTS) amb les claus de l'usuari.
 // Les claus viuen en aquest ordinador (~/.nio/config.json → voice) i mai no tornen a l'app.
 // L'àudio fa: mòbil → aquest ordinador → proveïdor. Proveïdors intercanviables: ElevenLabs i OpenAI.
+// Per entendre't també hi ha Whisper, gratis i sense clau, que funciona a l'ordinador mateix (whisper.js).
 const config = require('./config');
+const whisper = require('./whisper');
 
 const httpErr = (status, message) => Object.assign(new Error(message), { status });
 const MAX_AUDIO = 25 * 1024 * 1024;
@@ -67,11 +69,14 @@ async function errText(r) {
 }
 
 const cfgOf = (cfg) => cfg.voice || {};
-// Proveïdor per a cada cosa: el triat, si en té la clau; si no, qualsevol que en tingui.
+// Proveïdor per a cada cosa: el triat, si en té la clau (o Whisper, si és a punt); si no, Whisper
+// (gratis) per entendre't i, si tampoc, qualsevol que tingui clau.
 function pick(cfg, which) {
   const v = cfgOf(cfg), keys = v.keys || {};
   const want = v[which] || v.provider;
+  if (which === 'stt' && want === 'whisper' && whisper.ready(cfg)) return 'whisper';
   if (want && keys[want]) return want;
+  if (which === 'stt' && whisper.ready(cfg)) return 'whisper';
   return Object.keys(PROVIDERS).find((p) => keys[p]) || null;
 }
 
@@ -84,13 +89,15 @@ function status(cfg) {
     speak: !!v.speak,
     lang: v.lang || 'ca',
     voice: v.voice || {},
+    whisper: whisper.status(cfg),
   };
 }
 
 async function transcribe(cfg, audio, mime) {
   const p = pick(cfg, 'stt');
-  if (!p) throw httpErr(400, 'Aquest ordinador no té cap clau de veu. Posa-la a Ajustos → Veu.');
+  if (!p) throw httpErr(400, 'Aquest ordinador encara no té veu. Activa Whisper (gratis) o posa una clau a Ajustos → Veu.');
   if (!audio?.length) throw httpErr(400, 'No hi ha àudio');
+  if (p === 'whisper') return { text: await whisper.transcribe(cfg, audio, mime, cfgOf(cfg).lang || 'ca'), provider: p };
   return { text: (await PROVIDERS[p].stt(cfg.voice.keys[p], audio, mime, cfgOf(cfg).lang || 'ca')).trim(), provider: p };
 }
 
@@ -132,7 +139,8 @@ function routes(cfg) {
             v.keys[id] = k.trim();
           }
         }
-        for (const w of ['stt', 'tts']) if (body[w] === null || PROVIDERS[body[w]]) v[w] = body[w] || undefined;
+        for (const w of ['stt', 'tts']) if (body[w] === null || PROVIDERS[body[w]] || (w === 'stt' && body[w] === 'whisper')) v[w] = body[w] || undefined;
+        if (body.whisper && typeof body.whisper === 'object' && whisper.MODELS[body.whisper.model]) v.whisper = { ...v.whisper, model: body.whisper.model };
         if (typeof body.speak === 'boolean') v.speak = body.speak;
         if (typeof body.lang === 'string' && /^[a-z]{2}$/.test(body.lang)) v.lang = body.lang;
         if (body.voice && typeof body.voice === 'object') {
@@ -143,6 +151,16 @@ function routes(cfg) {
         return status(cfg);
       },
     },
+    {
+      // Whisper: «install» (Homebrew + model) o «download» (només el model).
+      method: 'POST',
+      path: '/api/voice/whisper',
+      handler: ({ body }) => {
+        if (body.action === 'install') whisper.install(cfg);
+        else whisper.download(cfg);
+        return status(cfg);
+      },
+    },
     {
       method: 'POST',
       path: '/api/voice/stt',

package/src/whisper.js

@@ -0,0 +1,148 @@
+'use strict';
+// Whisper a l'ordinador: transcripció gratuïta i sense clau. L'àudio no surt d'aquest ordinador.
+// Fa servir whisper.cpp (whisper-cli) amb ffmpeg per convertir l'àudio del navegador, o el
+// «whisper» de Python (openai-whisper) si és el que hi ha. El model de whisper.cpp es baixa una
+// sola vegada a ~/.nio/whisper.
+const fs = require('fs');
+const os = require('os');
+const path = require('path');
+const { execFile, spawn } = require('child_process');
+const { HOME } = require('./config');
+const { which, childEnv } = require('./proc');
+
+const DIR = path.join(HOME, 'whisper');
+const HF = 'https://huggingface.co/ggerganov/whisper.cpp/resolve/main';
+// Als Mac amb xip d'Apple, el turbo (quantitzat) va ràpid i entén molt bé el català. A la resta, el small.
+const MODELS = {
+  'large-v3-turbo-q5_0': { file: 'ggml-large-v3-turbo-q5_0.bin', size: 574e6, label: 'Turbo (millor)' },
+  'small-q5_1': { file: 'ggml-small-q5_1.bin', size: 190e6, label: 'Small (més lleuger)' },
+};
+const defaultModel = () => (process.platform === 'darwin' && process.arch === 'arm64' ? 'large-v3-turbo-q5_0' : 'small-q5_1');
+
+const job = { download: null, install: null, error: null }; // feines en curs (una de cada)
+
+const httpErr = (status, message) => Object.assign(new Error(message), { status });
+const cppBin = () => which('whisper-cli') || which('whisper-cpp');
+const pyBin = () => which('whisper');
+const modelId = (cfg) => (MODELS[cfg.voice?.whisper?.model] ? cfg.voice.whisper.model : defaultModel());
+const modelPath = (cfg) => path.join(DIR, MODELS[modelId(cfg)].file);
+
+function engine() {
+  if (cppBin() && which('ffmpeg')) return 'cpp';
+  if (pyBin() && which('ffmpeg')) return 'python';
+  return null;
+}
+
+function ready(cfg) {
+  const e = engine();
+  return e === 'python' || (e === 'cpp' && fs.existsSync(modelPath(cfg)));
+}
+
+function status(cfg) {
+  const e = engine(), missing = [];
+  if (!cppBin() && !pyBin()) missing.push('whisper-cpp');
+  if (!which('ffmpeg')) missing.push('ffmpeg');
+  return {
+    name: 'Whisper',
+    engine: e,
+    ready: ready(cfg),
+    missing,
+    model: modelId(cfg),
+    models: Object.fromEntries(Object.entries(MODELS).map(([id, m]) => [id, { label: m.label, size: m.size, have: fs.existsSync(path.join(DIR, m.file)) }])),
+    needsModel: e === 'cpp' && !fs.existsSync(modelPath(cfg)),
+    downloading: job.download ? { model: job.download.model, pct: job.download.pct } : null,
+    installing: !!job.install,
+    canInstall: process.platform === 'darwin' && !!which('brew'),
+    install: installHint(),
+    error: job.error,
+  };
+}
+
+function installHint() {
+  if (process.platform === 'darwin') return 'brew install whisper-cpp ffmpeg';
+  if (which('apt-get')) return 'sudo apt install ffmpeg pipx && pipx install openai-whisper';
+  return 'Instal·la whisper.cpp (whisper-cli) i ffmpeg';
+}
+
+// Baixa el model (una sola vegada). Torna de seguida; el progrés surt a status().
+function download(cfg) {
+  const id = modelId(cfg), m = MODELS[id], dest = path.join(DIR, m.file);
+  if (fs.existsSync(dest) || job.download) return;
+  fs.mkdirSync(DIR, { recursive: true });
+  const tmp = `${dest}.part`, d = (job.download = { model: id, pct: 0 });
+  job.error = null;
+  (async () => {
+    const res = await fetch(`${HF}/${m.file}`);
+    if (!res.ok || !res.body) throw new Error(`No he pogut baixar el model (${res.status})`);
+    const total = Number(res.headers.get('content-length')) || m.size;
+    const out = fs.createWriteStream(tmp);
+    let got = 0;
+    for await (const chunk of res.body) {
+      got += chunk.length;
+      d.pct = Math.min(99, Math.round((got / total) * 100));
+      if (!out.write(chunk)) await new Promise((r) => out.once('drain', r));
+    }
+    await new Promise((r, j) => out.end((e) => (e ? j(e) : r())));
+    fs.renameSync(tmp, dest);
+  })()
+    .catch((e) => { job.error = e.message; fs.rm(tmp, { force: true }, () => {}); })
+    .finally(() => { job.download = null; });
+}
+
+// Al Mac amb Homebrew, nio ho pot instal·lar sol. Després baixa el model.
+function install(cfg) {
+  if (job.install) return;
+  if (!(process.platform === 'darwin' && which('brew'))) throw httpErr(400, `Instal·la-ho al terminal de l’ordinador: ${installHint()}`);
+  job.error = null;
+  const want = ['whisper-cpp', 'ffmpeg'].filter((p) => (p === 'ffmpeg' ? !which('ffmpeg') : !cppBin()));
+  if (!want.length) return download(cfg);
+  const p = (job.install = spawn(which('brew'), ['install', ...want], { env: childEnv({ HOMEBREW_NO_AUTO_UPDATE: '1' }), stdio: ['ignore', 'ignore', 'pipe'] }));
+  let errOut = '';
+  p.stderr.on('data', (c) => { errOut = (errOut + c).slice(-400); });
+  p.on('error', (e) => { job.error = e.message; job.install = null; });
+  p.on('close', (code) => {
+    job.install = null;
+    if (code) job.error = `brew install ha fallat: ${errOut.trim().split('\n').pop() || code}`;
+    else download(cfg);
+  });
+}
+
+function run(bin, args, timeout = 180000) {
+  return new Promise((resolve, reject) => {
+    execFile(bin, args, { timeout, env: childEnv(), maxBuffer: 8 * 1024 * 1024 }, (err, stdout, stderr) => {
+      if (err) reject(httpErr(502, `Whisper: ${String(stderr || err.message).trim().split('\n').pop().slice(0, 200)}`));
+      else resolve(String(stdout));
+    });
+  });
+}
+
+function ext(mime) {
+  return /ogg/.test(mime || '') ? 'ogg' : /mp4|m4a|aac/.test(mime || '') ? 'm4a' : /mpeg|mp3/.test(mime || '') ? 'mp3' : /wav/.test(mime || '') ? 'wav' : 'webm';
+}
+
+async function transcribe(cfg, audio, mime, lang = 'ca') {
+  const e = engine();
+  if (!e) throw httpErr(400, `Falta Whisper en aquest ordinador. Instal·la’l: ${installHint()}`);
+  const tmp = fs.mkdtempSync(path.join(os.tmpdir(), 'nio-veu-'));
+  try {
+    const input = path.join(tmp, `audio.${ext(mime)}`);
+    fs.writeFileSync(input, audio);
+    if (e === 'python') {
+      await run(pyBin(), [input, '--model', 'turbo', '--language', lang, '--output_format', 'txt', '--output_dir', tmp, '--fp16', 'False', '--verbose', 'False']);
+      return fs.readFileSync(path.join(tmp, 'audio.txt'), 'utf8').trim();
+    }
+    if (!fs.existsSync(modelPath(cfg))) {
+      download(cfg);
+      throw httpErr(409, 'Baixant el model de Whisper (només el primer cop). Torna-ho a provar d’aquí a uns minuts.');
+    }
+    const wav = path.join(tmp, 'audio.wav');
+    await run(which('ffmpeg'), ['-nostdin', '-loglevel', 'error', '-i', input, '-ar', '16000', '-ac', '1', '-c:a', 'pcm_s16le', wav], 60000);
+    const threads = String(Math.max(1, Math.min(8, os.cpus().length - 1)));
+    const out = await run(cppBin(), ['-m', modelPath(cfg), '-f', wav, '-l', lang, '-t', threads, '-nt', '-np']);
+    return out.split('\n').map((l) => l.trim()).filter((l) => l && !/^\[(BLANK_AUDIO|Música|Music)\]$/i.test(l)).join(' ').trim();
+  } finally {
+    fs.rm(tmp, { recursive: true, force: true }, () => {});
+  }
+}
+
+module.exports = { status, ready, download, install, transcribe, MODELS, DIR };