// Speaker voiceprints: talks to the persistent spk_embed process (sherpa-onnx
// NeMo TitaNet embedding on CPU). embed(wav) → 192-dim vector; cosine match a
// diarization cluster against enrolled voiceprints → a name. All offline.
import { spawn } from 'node:child_process';
import fs from 'node:fs';
import readline from 'node:readline';

const BIN = '/opt/callrec/spk_embed';
const DIR = '/opt/sherpa/sherpa-onnx-v1.13.3-linux-aarch64-shared-cpu';
const EMB = process.env.DIAR_EMB || '/opt/sherpa/nemo_titanet_small.onnx';
export const THRESHOLD = +(process.env.SPK_THRESHOLD || 0.5);

let proc = null, ready = null, rl = null, pending = null, chain = Promise.resolve();

function start() {
  if (proc) return ready;
  if (!fs.existsSync(EMB)) return Promise.reject(new Error('Embedding-Modell fehlt: ' + EMB));
  proc = spawn('taskset', ['-c', '4-7', BIN, EMB],
    { env: { ...process.env, LD_LIBRARY_PATH: `${DIR}/lib` }, stdio: ['pipe', 'pipe', 'pipe'] });
  proc.stderr.on('data', d => process.stderr.write('[spk] ' + d));
  ready = new Promise((res, rej) => {
    const to = setTimeout(() => rej(new Error('spk READY timeout')), 60000);
    rl = readline.createInterface({ input: proc.stdout });
    rl.on('line', line => {
      if (line.startsWith('READY')) { clearTimeout(to); res(); return; }
      if (!pending) return;
      const p = pending; pending = null;
      if (line.startsWith('OK ')) p.resolve(line.slice(3).split(',').map(Number));
      else p.reject(new Error(line.replace(/^ERR\s*/, '') || 'spk error'));
    });
  });
  const cleanup = () => { proc = null; ready = null; rl = null; if (pending) { pending.reject(new Error('spk beendet')); pending = null; } };
  proc.on('exit', cleanup); proc.on('error', cleanup);
  return ready;
}

// Extract the voiceprint for a 16 kHz mono wav file.
export function embed(wavPath) {
  chain = chain.then(async () => {
    await start();
    return await new Promise((res, rej) => { pending = { resolve: res, reject: rej }; proc.stdin.write(wavPath + '\n'); });
  });
  return chain;
}

export function cosine(a, b) {
  let d = 0, na = 0, nb = 0;
  for (let i = 0; i < a.length; i++) { d += a[i] * b[i]; na += a[i] * a[i]; nb += b[i] * b[i]; }
  return d / (Math.sqrt(na) * Math.sqrt(nb) + 1e-9);
}

// Best enrolled voiceprint for an embedding, or null if below threshold.
export function bestMatch(emb, voiceprints, threshold = THRESHOLD) {
  let best = null, bs = -1;
  for (const v of voiceprints) { const s = cosine(emb, v.vec); if (s > bs) { bs = s; best = v; } }
  return (best && bs >= threshold) ? { id: best.id, name: best.name, score: bs } : null;
}

export function status() { return { model: EMB, present: fs.existsSync(EMB), running: !!proc, threshold: THRESHOLD }; }