server.js

import express from 'express';
import cors from 'cors';
import ytdl from '@distube/ytdl-core';
import yts from 'youtube-yts';
import { HttpsProxyAgent } from 'https-proxy-agent';
import { exec } from 'child_process';
import { promisify } from 'util';
import fs from 'fs';
import path from 'path';

const execAsync = promisify(exec);
const COOKIES_PATH = path.resolve('./cookies.txt');

// Dynamically write cookies from environment secret if present
if (process.env.YT_COOKIES) {
  console.log('[Cookies] Dynamically writing cookies.txt from environment variable...');
  try {
    fs.writeFileSync(COOKIES_PATH, process.env.YT_COOKIES, 'utf-8');
    console.log('[Cookies] Successfully wrote cookies.txt to', COOKIES_PATH);
  } catch (err) {
    console.error('[Cookies] Failed to write cookies.txt:', err.message);
  }
} else {
  console.log('[Cookies] YT_COOKIES environment secret not set, checking for local cookies.txt file...');
}

const app = express();
app.use(cors());
app.use(express.json());

const PORT = process.env.PORT || 7860;

// High-Availability Cobalt Pool
const COBALT_INSTANCES = [
  "https://cobalt.omega.wolfy.love",
  "https://apicobalt.mgytr.top",
  "https://api.qwkuns.me",
  "https://subito-c.meowing.de",
  "https://cobalt.alpha.wolfy.love"
];

// LAYER 4: Invidious Proxy Fallback (Ultimate 429/Age-Gate Bypass)
const INVIDIOUS_INSTANCES = [
  "https://invidious.f5.si",
  "https://invidious.tiekoetter.com",
  "https://invidious.nerdvpn.de",
  "https://yt.chocolatemoo53.com",
  "https://inv.thepixora.com",
  "https://inv.nadeko.net"
];

// LAYER 5: Piped API Fallback (Independent of YouTube's auth challenges)
const PIPED_INSTANCES = [
  "https://pipedapi.kavin.rocks",
  "https://pipedapi.adminforge.de",
  "https://piped-api.lunar.icu",
  "https://api.piped.projectsegfau.lt",
  "https://pipedapi.r4fo.com",
  "https://pipedapi.darkness.services"
];

// Innertube clients raced in parallel — same approach as the CF worker.
// ANDROID_MUSIC is fastest; ANDROID_VR and TVHTML5 are independent fallbacks
// that historically bypass the "Sign in" challenge that hits ANDROID_MUSIC.
const INNERTUBE_CLIENTS = [
  { name: "ANDROID_MUSIC", version: "6.19.52", userAgent: "Mozilla/5.0 (Linux; Android 10; K) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Mobile Safari/537.36" },
  { name: "ANDROID_VR",    version: "1.55.18", userAgent: "Mozilla/5.0 (Linux; Android 10; Quest 2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Mobile Safari/537.36" },
  { name: "TVHTML5_SIMPLY_EMBEDDED_PLAYER", version: "2.0", userAgent: "Mozilla/5.0 (PlayStation; PlayStation 4/12.00) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/16.0 Safari/605.1.15" }
];

// ============================================================
// LAYER 0 (PRIMARY): IZGIV-Botto 3rd-Party API Cascade
// Ported from IZGIV-Botto/commands/song.js + play.js — these
// downloader APIs (EliteProTech, Yupra, Okatsu, apis-keith)
// are battle-tested in production and bypass every
// signatureCipher / "Sign in" / 403 challenge. Raced in
// parallel with 3-attempt retry (1s/2s/3s backoff) per
// the IZGIV-Botto tryRequest pattern.
// ============================================================

// Sanity check the upstream audio URL before returning it.
// Some 3rd-party APIs (notably EliteProTech right now) return URLs to
// well-formed MP4 headers but with empty/zeroed `mdat` boxes — the
// file looks valid but plays silence. A HEAD probe of Content-Length
// against a sane floor (100KB — a real 3-min song at 64kbps is ~1.4MB)
// catches this without a full body download.
const MIN_AUDIO_BYTES = 100_000;
async function validateAudioSize(url, label) {
  try {
    const head = await fetchWithTimeout(url, {
      method: 'HEAD',
      headers: { 'User-Agent': 'Mozilla/5.0' },
      timeout: 5000
    });
    if (!head.ok) throw new Error(`HEAD status ${head.status}`);
    const cl = parseInt(head.headers.get('content-length') || '0', 10);
    if (!cl || cl < MIN_AUDIO_BYTES) {
      throw new Error(`Too-small file: ${cl} bytes (min ${MIN_AUDIO_BYTES})`);
    }
    return true;
  } catch (err) {
    console.warn(`[Izgiv:${label}] size validation failed: ${err.message}`);
    throw err;
  }
}

async function getEliteProTechUrl(videoUrl) {
  const apiUrl = `https://eliteprotech-apis.zone.id/ytdown?url=${encodeURIComponent(videoUrl)}&format=mp3`;
  const res = await fetchWithTimeout(apiUrl, {
    headers: { 'User-Agent': 'Mozilla/5.0' },
    timeout: 6000
  });
  if (!res.ok) throw new Error(`Status ${res.status}`);
  const data = await res.json();
  // EliteProTech shape: { success: true, title, downloadURL } | { success: false, ... }
  if (!data?.success || !data?.downloadURL) throw new Error('No downloadURL in response');
  await validateAudioSize(data.downloadURL, 'EliteProTech');
  console.log(`[Izgiv:EliteProTech] ✅ resolved (title="${data.title}")`);
  return { url: data.downloadURL, title: data.title };
}

async function getYupraUrl(videoUrl) {
  const apiUrl = `https://api.yupra.my.id/api/downloader/ytmp3?url=${encodeURIComponent(videoUrl)}`;
  const res = await fetchWithTimeout(apiUrl, {
    headers: { 'User-Agent': 'Mozilla/5.0' },
    timeout: 6000
  });
  if (!res.ok) throw new Error(`Status ${res.status}`);
  const data = await res.json();
  // Yupra shape: { success: true, data: { title, thumbnail, download_url, ... } }
  if (!data?.success || !data?.data?.download_url) throw new Error('No download_url in response');
  await validateAudioSize(data.data.download_url, 'Yupra');
  console.log(`[Izgiv:Yupra] ✅ resolved (title="${data.data.title}")`);
  return { url: data.data.download_url, title: data.data.title, thumbnail: data.data.thumbnail };
}

async function getOkatsuUrl(videoUrl) {
  const apiUrl = `https://okatsu-rolezapiiz.vercel.app/downloader/ytmp3?url=${encodeURIComponent(videoUrl)}`;
  const res = await fetchWithTimeout(apiUrl, {
    headers: { 'User-Agent': 'Mozilla/5.0' },
    timeout: 6000
  });
  if (res.status === 451) {
    console.warn(`[Izgiv:Okatsu] ⛔ 451 region/copyright block — skipping API`);
    return { _skipped: true };
  }
  if (!res.ok) throw new Error(`Status ${res.status}`);
  const data = await res.json();
  // Okatsu audio shape: { status, creator, title, format, thumb, duration, cached, dl }
  if (!data?.dl) throw new Error('No dl URL in response');
  await validateAudioSize(data.dl, 'Okatsu');
  console.log(`[Izgiv:Okatsu] ✅ resolved (title="${data.title}")`);
  return { url: data.dl, title: data.title, thumbnail: data.thumb };
}

async function getApisKeithUrl(videoUrl) {
  const apiUrl = `https://apis-keith.vercel.app/download/dlmp3?url=${encodeURIComponent(videoUrl)}`;
  const res = await fetchWithTimeout(apiUrl, {
    headers: { 'User-Agent': 'Mozilla/5.0' },
    timeout: 6000
  });
  if (!res.ok) throw new Error(`Status ${res.status}`);
  const data = await res.json();
  // ApisKeith shape: { status: true, result: { title, mp3, image, duration, quality } }
  if (!data?.result?.mp3) throw new Error('No mp3 URL in response');
  await validateAudioSize(data.result.mp3, 'ApisKeith');
  console.log(`[Izgiv:ApisKeith] ✅ resolved (title="${data.result.title}")`);
  return { url: data.result.mp3, title: data.result.title, thumbnail: data.result.image };
}

async function getA2ZConverterUrl(videoUrl) {
  const apiUrl = `https://www.a2zconverter.com/api/get-proxy-data?url=${encodeURIComponent(videoUrl)}`;
  const res = await fetchWithTimeout(apiUrl, {
    headers: {
      'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36',
      'Referer': 'https://www.a2zconverter.com/youtube-video-downloader'
    },
    timeout: 6000
  });
  if (!res.ok) throw new Error(`Status ${res.status}`);
  const data = await res.json();
  if (!data?.status || !data?.data?.items?.audio) {
    throw new Error('Invalid or failed response from a2zconverter');
  }
  const audios = data.data.items.audio;
  if (!audios || audios.length === 0) {
    throw new Error('No audio streams found in a2zconverter data');
  }
  const bestAudio = audios[0];
  if (!bestAudio.url) {
    throw new Error('No audio URL in a2zconverter item');
  }
  await validateAudioSize(bestAudio.url, 'A2ZConverter');
  console.log(`[Izgiv:A2ZConverter] ✅ resolved (title="${data.data.title}")`);
  return { url: bestAudio.url, title: data.data.title, thumbnail: data.data.cover };
}

async function tryRequest(getter, label) {
  try {
    const data = await getter();
    if (data && data._skipped) return { skipped: true };
    if (data && data.url) return data;
  } catch (err) {
    console.warn(`[Izgiv:${label}] failed: ${err.message}`);
  }
  return null;
}

async function resolveIzgivBotto(videoId) {
  const videoUrl = `https://www.youtube.com/watch?v=${videoId}`;
  console.log(`[Izgiv] Racing all 5 APIs in parallel for ${videoId}...`);
  try {
    const winner = await Promise.any([
      tryRequest(() => getEliteProTechUrl(videoUrl), 'EliteProTech').then(res => {
        if (res && res.url) {
          res.strategy = 'izgiv:eliteprotech';
          return res;
        }
        throw new Error('EliteProTech returned empty');
      }),
      tryRequest(() => getYupraUrl(videoUrl), 'Yupra').then(res => {
        if (res && res.url) {
          res.strategy = 'izgiv:yupra';
          return res;
        }
        throw new Error('Yupra returned empty');
      }),
      tryRequest(() => getOkatsuUrl(videoUrl), 'Okatsu').then(res => {
        if (res && res.url) {
          res.strategy = 'izgiv:okatsu';
          return res;
        }
        throw new Error('Okatsu returned empty');
      }),
      tryRequest(() => getApisKeithUrl(videoUrl), 'ApisKeith').then(res => {
        if (res && res.url) {
          res.strategy = 'izgiv:apis-keith';
          return res;
        }
        throw new Error('ApisKeith returned empty');
      }),
      tryRequest(() => getA2ZConverterUrl(videoUrl), 'A2ZConverter').then(res => {
        if (res && res.url) {
          res.strategy = 'izgiv:a2zconverter';
          return res;
        }
        throw new Error('A2ZConverter returned empty');
      })
    ]);
    console.log(`[Izgiv] ✅ Resolved ${videoId} via ${winner.strategy}`);
    return winner;
  } catch (err) {
    console.warn('[Izgiv] All IZGIV APIs failed — falling through to next Tier');
    return null;
  }
}

// --- HIGH-PERFORMANCE OUTBOUND PROXY POOL ROTATION ---
const PROXY_POOL = process.env.PROXY_POOL ? process.env.PROXY_POOL.split(',').map(p => p.trim()) : [];
let proxyIndex = 0;

function getNextProxy() {
  if (PROXY_POOL.length === 0) return null;
  const proxy = PROXY_POOL[proxyIndex];
  proxyIndex = (proxyIndex + 1) % PROXY_POOL.length;
  return proxy;
}

// Gold-Standard Fetch Timeout Utility with Proxy Rotation
async function fetchWithTimeout(url, options = {}) {
  const { timeout = 5000, ...fetchOptions } = options;
  const controller = new AbortController();
  const id = setTimeout(() => controller.abort(), timeout);
  
  // Dynamic proxy injection
  const proxy = getNextProxy();
  if (proxy && !fetchOptions.agent) {
    console.log(`[Proxy Rotator] Directing request through: ${proxy}`);
    fetchOptions.agent = new HttpsProxyAgent(proxy);
  }

  try {
    const response = await fetch(url, {
      ...fetchOptions,
      signal: controller.signal
    });
    clearTimeout(id);
    return response;
  } catch (err) {
    clearTimeout(id);
    throw err;
  }
}

// Memory cache for resolved stream URLs to handle range/concurrency requests
const streamCache = new Map();
const CACHE_TTL_MS = 120 * 60 * 1000; // 120 minutes (2 hours) to minimize upstream API quota usage

// ── Bot-side Telegram-channel cache (LONG-TERM, persistent) ────────────────
// The AriseRobot bot runs a small aiohttp sidecar that maps yt_id → a
// Cloudflare-R2-hosted MP3 (mirrored from the bot's Telegram cache
// channel). When a song is cached, we can serve the R2 URL directly and
// skip the entire 5-layer resolution cascade. Hits are essentially free;
// misses fall through to the cascade unchanged.
const BOT_CACHE_URL = (process.env.CACHE_SERVER_URL || 'http://127.0.0.1:7861').replace(/\/$/, '');
const BOT_CACHE_TOKEN = process.env.CACHE_SERVER_TOKEN || 'arise-music-cache';
const BOT_CACHE_TIMEOUT_MS = parseInt(process.env.CACHE_SERVER_TIMEOUT_MS || '2500', 10);

async function checkBotCache(videoId) {
  if (!videoId) return null;
  const url = `${BOT_CACHE_URL}/cache/lookup?yt_id=${encodeURIComponent(videoId)}`;
  const controller = new AbortController();
  const id = setTimeout(() => controller.abort(), BOT_CACHE_TIMEOUT_MS);
  try {
    const res = await fetch(url, {
      method: 'GET',
      headers: { Authorization: `Bearer ${BOT_CACHE_TOKEN}` },
      signal: controller.signal,
    });
    if (res.status === 404) return null;
    if (!res.ok) {
      console.warn(`[BotCache] lookup ${videoId} status=${res.status}`);
      return null;
    }
    const data = await res.json();
    if (data && data.hit) {
      console.log(`[BotCache] ✅ HIT for ${videoId} (msg=${data.messageId})`);
      return data;
    }
    return null;
  } catch (err) {
    // Bot not running, slow, or unreachable → fall through to cascade.
    if (err.name !== 'AbortError') {
      console.warn(`[BotCache] lookup ${videoId} failed: ${err.message}`);
    }
    return null;
  } finally {
    clearTimeout(id);
  }
}

// ── Bot-cache audio proxy ──────────────────────────────────────────────────
// Streams the cached audio from the bot's /cache/audio endpoint and
// re-serves it to the public. The node_proxy becomes the streaming
// hot path — that's the trade-off for not using an external CDN like R2.
//
// In-memory LRU (~50 MB cap) avoids hitting the bot for repeat
// requests of the same track. The bot already keeps its own on-disk
// copy in scratch/music_cache_readout/, so this is just a second-tier
// cache to absorb the bursty Range-request pattern from <audio>.

const AUDIO_CACHE = new Map();
const AUDIO_CACHE_CAP = 50 * 1024 * 1024; // 50 MB total
let audioCacheBytes = 0;

async function proxyBotAudio(videoId, req, res) {
  // LRU: refresh key order on access
  if (AUDIO_CACHE.has(videoId)) {
    const entry = AUDIO_CACHE.get(videoId);
    AUDIO_CACHE.delete(videoId);
    AUDIO_CACHE.set(videoId, entry);
    return sendCachedAudio(entry, req, res);
  }

  const url = `${BOT_CACHE_URL}/cache/audio?yt_id=${encodeURIComponent(videoId)}`;
  const fetchHeaders = {};
  if (req.headers.range) {
    fetchHeaders['Range'] = req.headers.range;
  }
  const controller = new AbortController();
  const id = setTimeout(() => controller.abort(), 30_000);

  let upstream;
  try {
    upstream = await fetch(url, {
      method: 'GET',
      headers: { Authorization: `Bearer ${BOT_CACHE_TOKEN}`, ...fetchHeaders },
      signal: controller.signal,
    });
  } catch (err) {
    clearTimeout(id);
    console.warn(`[BotCache] audio fetch ${videoId} failed: ${err.message}`);
    if (!res.headersSent) {
      res.status(502).json({ error: `Bot cache unreachable: ${err.message}` });
    }
    return;
  }
  clearTimeout(id);

  if (!upstream.ok && upstream.status !== 206) {
    if (upstream.status === 404) {
      console.log(`[BotCache] audio ${videoId} → 404 (miss)`);
    } else {
      console.warn(`[BotCache] audio ${videoId} → ${upstream.status}`);
    }
    const body = await upstream.text().catch(() => '');
    res.status(upstream.status).json({ error: body || `Bot cache ${upstream.status}` });
    return;
  }

  const contentType = upstream.headers.get('content-type') || 'audio/mpeg';
  const contentLength = parseInt(upstream.headers.get('content-length') || '0', 10);

  // Build a tiny CORS-friendly header bag
  const headers = {
    'Content-Type': contentType,
    'Accept-Ranges': 'bytes',
    'Cache-Control': 'public, max-age=3600',
    'Access-Control-Allow-Origin': '*',
    'Access-Control-Allow-Methods': 'GET, OPTIONS',
    'Access-Control-Allow-Headers': 'Range, Content-Type',
    'Access-Control-Expose-Headers': 'Content-Length, Content-Range, Accept-Ranges',
  };
  for (const h of ['content-length', 'content-range']) {
    const v = upstream.headers.get(h);
    if (v) headers[h] = v;
  }

  res.writeHead(upstream.status, headers);

  // Full-body cache: only when the response is small enough to keep in
  // the in-memory budget. We buffer, then either send or set up streaming
  // with an "on close" cache write.
  if (contentLength > 0 && contentLength <= 8 * 1024 * 1024) {
    try {
      const ab = await upstream.arrayBuffer();
      const buf = Buffer.from(ab);
      res.end(buf);

      // Promote to LRU cache if there's still room
      evictIfNeeded(contentLength);
      if (audioCacheBytes + contentLength <= AUDIO_CACHE_CAP) {
        AUDIO_CACHE.set(videoId, { buf, contentType });
        audioCacheBytes += contentLength;
      }
      return;
    } catch (err) {
      if (!res.headersSent) {
        res.status(502).json({ error: `Bot cache read failed: ${err.message}` });
      }
      return;
    }
  }

  // Large response: stream straight through (no in-memory cache).
  const reader = upstream.body.getReader();
  const push = async () => {
    while (true) {
      const { done, value } = await reader.read();
      if (done) break;
      res.write(value);
    }
    res.end();
  };
  push().catch(err => {
    console.warn(`[BotCache] stream ${videoId} error: ${err.message}`);
    res.end();
  });
  req.on('close', () => reader.cancel());
}

function sendCachedAudio(entry, req, res) {
  const headers = {
    'Content-Type': entry.contentType || 'audio/mpeg',
    'Content-Length': String(entry.buf.length),
    'Accept-Ranges': 'bytes',
    'Cache-Control': 'public, max-age=3600',
    'Access-Control-Allow-Origin': '*',
    'Access-Control-Allow-Headers': 'Range, Content-Type',
    'Access-Control-Expose-Headers': 'Content-Length, Content-Range, Accept-Ranges',
  };

  // Honor Range for already-buffered responses
  if (req.headers.range && entry.buf) {
    const m = /^bytes=(\d+)-(\d+)?$/.exec(req.headers.range);
    if (m) {
      const start = parseInt(m[1], 10);
      const end = m[2] ? parseInt(m[2], 10) : entry.buf.length - 1;
      if (start <= end && end < entry.buf.length) {
        const slice = entry.buf.subarray(start, end + 1);
        headers['Content-Range'] = `bytes ${start}-${end}/${entry.buf.length}`;
        headers['Content-Length'] = String(slice.length);
        res.writeHead(206, headers);
        res.end(slice);
        return;
      }
    }
  }

  res.writeHead(200, headers);
  res.end(entry.buf);
}

function evictIfNeeded(incoming) {
  while (audioCacheBytes + incoming > AUDIO_CACHE_CAP && AUDIO_CACHE.size > 0) {
    const oldestKey = AUDIO_CACHE.keys().next().value;
    const oldest = AUDIO_CACHE.get(oldestKey);
    if (!oldest) break;
    audioCacheBytes -= oldest.buf.length;
    AUDIO_CACHE.delete(oldestKey);
  }
}

// Health check
app.get('/health', (req, res) => {
  try {
    res.status(200).json({
      status: 'healthy',
      service: 'node-ytdl-proxy',
      timestamp: new Date().toISOString()
    });
  } catch (err) {
    res.status(500).json({ error: err.message });
  }
});

// Public cache audio proxy. Streams from the bot's /cache/audio endpoint
// over loopback. The node_proxy is the public-facing endpoint; the bot
// stays on a private network. Hit LRU is 50 MB total.
app.get('/cache/audio', async (req, res) => {
  const ytId = (req.query.yt_id || '').trim();
  if (!ytId) {
    return res.status(400).json({ error: 'Missing required query parameter "yt_id"' });
  }
  return proxyBotAudio(ytId, req, res);
});

// LAYER 1: Multi-Client Innertube API Extractor — race ANDROID_MUSIC + ANDROID_VR + TVHTML5
// Returns the first direct audio URL any client gives. Skips signatureCipher entries
// (those require n/sig deciphering which the yt-dlp/pytubefix fallback handles).
async function innertubeFetchOnce(videoId, client) {
  const url = "https://www.youtube.com/youtubei/v1/player";
  const payload = {
    videoId,
    context: {
      client: {
        clientName: client.name,
        clientVersion: client.version,
        hl: "en",
        gl: "US"
      }
    },
    contentCheckOk: true,
    racyCheckOk: true
  };

  try {
    const response = await fetchWithTimeout(url, {
      method: "POST",
      headers: {
        "Content-Type": "application/json",
        "User-Agent": client.userAgent
      },
      body: JSON.stringify(payload),
      timeout: 5000
    });

    if (!response.ok) {
      console.warn(`[Innertube:${client.name}] status ${response.status}`);
      return null;
    }

    const data = await response.json();
    const formats = data?.streamingData?.adaptiveFormats || [];
    const directFormats = formats.filter(f => f.url && !f.signatureCipher);

    // Prefer format 140 (128kbps AAC m4a) — broadest device compatibility
    let bestAudio = directFormats.find(f => f.mimeType?.includes("audio/mp4") && f.itag === 140)
      || directFormats.find(f => f.mimeType?.includes("audio/mp4"))
      || directFormats.find(f => f.mimeType?.includes("audio/webm"))
      || directFormats.find(f => f.mimeType?.includes("audio/"));

    if (bestAudio && bestAudio.url) {
      console.log(`[Innertube:${client.name}] ✅ direct stream for ${videoId}`);
      return { url: bestAudio.url, contentLength: bestAudio.contentLength };
    }
    return null;
  } catch (err) {
    console.warn(`[Innertube:${client.name}] failed: ${err.message}`);
    return null;
  }
}

async function extractDirectStreamUrl(videoId) {
  console.log(`[Innertube] Racing ${INNERTUBE_CLIENTS.length} clients for: ${videoId}`);
  const attempts = INNERTUBE_CLIENTS.map(c => innertubeFetchOnce(videoId, c));
  const results = await Promise.all(attempts);
  const winner = results.find(r => r && r.url);
  if (winner) {
    return winner;
  }
  console.warn("[Innertube] All clients failed");
  return null;
}

// Spawn pytubefix Python extractor
async function resolvePytubefix(videoId) {
  try {
    console.log(`[Pytubefix] Spawning pytubefix extraction for: ${videoId}`);
    const cmd = `python3 resolve_pytubefix.py "${videoId}"`;
    const { stdout } = await execAsync(cmd, { timeout: 6000 });
    const data = JSON.parse(stdout);
    if (data && data.url) {
      console.log(`[Pytubefix] Successfully resolved stream!`);
      return {
        url: data.url,
        contentLength: data.contentLength || null
      };
    }
    if (data && data.error) {
      console.warn(`[Pytubefix] Extraction warning: ${data.error}`);
    }
    return null;
  } catch (err) {
    console.error(`[Pytubefix] Execution failed: ${err.message}`);
    return null;
  }
}

// LAYER 2: Double-Engine CLI Extractor (Pytubefix Stage 2.1 & yt-dlp Stage 2.2)
async function resolveYtdl(videoId) {
  // Stage 2.1: Try pytubefix First (Extremely robust & native JS-free decryption)
  const pytubeResult = await resolvePytubefix(videoId);
  if (pytubeResult) return pytubeResult;

  // Stage 2.2: Fall back to direct yt-dlp CLI
  const videoUrl = `https://www.youtube.com/watch?v=${videoId}`;
  try {
    console.log(`[yt-dlp CLI] Stage 2.2: Falling back to direct yt-dlp extraction for: ${videoId}`);
    
    let cmd = `yt-dlp -j --format "140/ba/best" --socket-timeout 10 --no-check-certificate --legacy-server-connect --remote-components ejs:github "${videoUrl}"`;
    if (fs.existsSync(COOKIES_PATH)) {
      cmd += ` --cookies "${COOKIES_PATH}"`;
    }
    
    const { stdout } = await execAsync(cmd, { timeout: 15000 });
    const data = JSON.parse(stdout);
    if (data && data.url) {
      console.log(`[yt-dlp CLI] Stage 2.2: DIRECT extraction succeeded!`);
      return {
        url: data.url,
        contentLength: data.filesize || data.filesize_approx || null
      };
    }
    return null;
  } catch (err) {
    console.error(`[yt-dlp CLI] Stage 2.2: DIRECT extraction failed: ${err.message}`);
    return null;
  }
}

// LAYER 3: Cobalt API Fallback — race ALL instances in parallel, first success wins
async function resolveCobalt(videoId) {
  const promises = COBALT_INSTANCES.map(async (api) => {
    try {
      console.log(`[Cobalt] Attempting resolution via: ${api} for: ${videoId}`);
      const res = await fetchWithTimeout(`${api}/api/json`, {
        method: 'POST',
        headers: {
          'Accept': 'application/json',
          'Content-Type': 'application/json',
          'User-Agent': 'Mozilla/5.0'
        },
        body: JSON.stringify({
          url: `https://www.youtube.com/watch?v=${videoId}`,
          downloadMode: 'audio',
          audioFormat: 'best'
        }),
        timeout: 5000
      });
      if (res.ok) {
        const data = await res.json();
        if (data && data.url) {
          console.log(`[Cobalt] Successfully resolved stream from ${api}`);
          return { url: data.url };
        }
      }
      throw new Error(`Status ${res.status}`);
    } catch (err) {
      console.warn(`[Cobalt] API ${api} failed: ${err.message}`);
      throw err;
    }
  });
  try {
    return await Promise.any(promises);
  } catch (err) {
    console.warn(`[Cobalt] All ${COBALT_INSTANCES.length} instances exhausted.`);
    return null;
  }
}

// LAYER 4: Invidious Proxy Fallback — race ALL instances in parallel
async function resolveInvidious(videoId) {
  const promises = INVIDIOUS_INSTANCES.map(async (instance) => {
    const streamUrl = `${instance}/latest_version?id=${videoId}&itag=140&local=true`;
    try {
      console.log(`[Invidious] Attempting proxy stream check via: ${streamUrl}`);
      const resp = await fetchWithTimeout(streamUrl, {
        method: 'HEAD',
        timeout: 5000
      });
      if (resp.ok && resp.headers.get('content-type')?.includes('audio')) {
        console.log(`[Invidious] Successfully found active stream source: ${streamUrl}`);
        return {
          url: streamUrl,
          contentLength: resp.headers.get('content-length')
        };
      }
      throw new Error('Bad content-type or status');
    } catch (err) {
      console.warn(`[Invidious] Instance ${instance} failed: ${err.message}`);
      throw err;
    }
  });
  try {
    return await Promise.any(promises);
  } catch (err) {
    console.warn(`[Invidious] All ${INVIDIOUS_INSTANCES.length} instances exhausted.`);
    return null;
  }
}

// LAYER 5: Piped API Fallback — race ALL Piped instances, return lowest-bitrate audio stream
async function resolvePiped(videoId) {
  const promises = PIPED_INSTANCES.map(async (instance) => {
    try {
      console.log(`[Piped] Resolving via: ${instance}/streams/${videoId}`);
      const resp = await fetchWithTimeout(`${instance}/streams/${videoId}`, {
        headers: { 'User-Agent': 'Mozilla/5.0' },
        timeout: 5000
      });
      if (!resp.ok) throw new Error(`Status ${resp.status}`);
      const data = await resp.json();
      const streams = data?.audioStreams || [];
      if (streams.length === 0) throw new Error('No audio streams');
      // Pick the lowest bitrate for fastest first-byte to the audio element
      const sorted = [...streams].sort((a, b) => (a.bitrate || 0) - (b.bitrate || 0));
      const audio = sorted.find(s => s.mimeType?.includes('audio/mp4')) || sorted[0];
      if (!audio?.url) throw new Error('No audio URL in response');
      console.log(`[Piped] ✅ Resolved from ${instance}`);
      return { url: audio.url, contentLength: null };
    } catch (err) {
      console.warn(`[Piped] ${instance} failed: ${err.message}`);
      throw err;
    }
  });
  try {
    return await Promise.any(promises);
  } catch (err) {
    console.warn(`[Piped] All ${PIPED_INSTANCES.length} instances exhausted.`);
    return null;
  }
}

// Helper to fetch robust YouTube Metadata via YouTube's official, unblocked oEmbed API
async function getVideoMetadata(videoId) {
  try {
    console.log(`[Metadata] Fetching oEmbed for video: ${videoId}`);
    const oembedUrl = `https://www.youtube.com/oembed?url=https://www.youtube.com/watch?v=${videoId}&format=json`;
    const resp = await fetchWithTimeout(oembedUrl, { timeout: 3000 });
    if (resp.ok) {
      const data = await resp.json();
      return {
        title: data.title || 'YouTube Audio Feed',
        duration: 0,
        thumbnail: data.thumbnail_url || `https://i.ytimg.com/vi/${videoId}/hqdefault.jpg`,
        artist: data.author_name || 'Various Artists'
      };
    }
  } catch (err) {
    console.warn(`[Metadata] oEmbed lookup failed: ${err.message}`);
  }

  // Safe fallback (avoid youtube-yts scraping as it gets blocked/hangs in cloud environments)
  return {
    title: 'YouTube Audio Feed',
    duration: 0,
    thumbnail: `https://i.ytimg.com/vi/${videoId}/hqdefault.jpg`,
    artist: 'Various Artists'
  };
}

// Master Resolver Orchestrator — runs the cheaper/faster strategies in a tiered cascade,
// prioritizing high-quality direct CDN URLs and falling back sequentially to avoid spamming subrequests.
async function masterResolve(videoId) {
  // Group 1: High-Quality Direct APIs raced in parallel (IZGIV, Innertube, Cobalt)
  try {
    console.log(`[Resolve] Group 1: Racing IZGIV, Innertube, and Cobalt in parallel for ${videoId}...`);
    const res = await Promise.any([
      resolveIzgivBotto(videoId).then(r => {
        if (r && r.url) return r;
        throw new Error('Izgiv failed');
      }),
      extractDirectStreamUrl(videoId).then(r => {
        if (r && r.url) return r;
        throw new Error('Innertube failed');
      }),
      resolveCobalt(videoId).then(r => {
        if (r && r.url) return r;
        throw new Error('Cobalt failed');
      })
    ]);
    if (res && res.url) {
      console.log(`[Resolve] ✅ Group 1 resolved stream for ${videoId}`);
      return res;
    }
  } catch (err) {
    console.warn(`[Resolve] Group 1 failed for ${videoId}: ${err.message}. Trying Group 2...`);
  }

  // Group 2: Public Pools raced in parallel (Piped, Invidious)
  try {
    console.log(`[Resolve] Group 2: Racing Piped and Invidious in parallel for ${videoId}...`);
    const res = await Promise.any([
      resolvePiped(videoId).then(r => {
        if (r && r.url) return r;
        throw new Error('Piped failed');
      }),
      resolveInvidious(videoId).then(r => {
        if (r && r.url) return r;
        throw new Error('Invidious failed');
      })
    ]);
    if (res && res.url) {
      console.log(`[Resolve] ✅ Group 2 resolved stream for ${videoId}`);
      return res;
    }
  } catch (err) {
    console.warn(`[Resolve] Group 2 failed for ${videoId}: ${err.message}`);
  }

  return null;
}

// Search YouTube
app.get('/api/search', async (req, res) => {
  const query = req.query.q;
  if (!query) {
    return res.status(400).json({ error: 'Missing search query parameter "q"' });
  }

  try {
    console.log(`🔎 Searching YouTube for: ${query}`);
    const searchResult = await yts({ query, hl: 'en', gl: 'US' });
    const videos = (searchResult.videos || []).slice(0, 10).map(video => ({
      id: video.videoId,
      title: video.title,
      artist: video.author.name,
      coverUrl: video.thumbnail || video.image,
      duration: video.seconds,
      url: video.url
    }));

    console.log(`✅ Found ${videos.length} search results`);
    res.status(200).json({ results: videos });
  } catch (err) {
    console.error('❌ Search failed:', err);
    res.status(500).json({ error: err.message });
  }
});

// Resolve Stream URL
app.get('/api/resolve', async (req, res) => {
  try {
    const videoId = req.query.v || req.query.url;
    if (!videoId) {
      return res.status(400).json({ error: 'Missing parameter "v" or "url"' });
    }

    const cleanId = videoId.startsWith('http') ? ytdl.getVideoID(videoId) : videoId;

    // Check if cached
    const cached = streamCache.get(cleanId);
    if (cached && Date.now() < cached.expiresAt) {
      console.log(`🎯 Cache HIT for resolve: ${cleanId}`);
      const meta = await getVideoMetadata(cleanId).catch(() => ({}));
      return res.status(200).json({
        url: cached.url,
        title: meta.title || 'YouTube Audio Feed',
        duration: meta.duration || 0,
        thumbnail: meta.thumbnail || `https://i.ytimg.com/vi/${cleanId}/hqdefault.jpg`,
        artist: meta.artist || 'Various Artists',
        format: 'm4a',
        source: cached.source || 'memory'
      });
    }

    // ── LAYER -1 (PRIMARY): Bot-side Telegram-channel cache ──
    // The bot maintains a yt_id → Telegram-channel → /cache/audio cache.
    // On hit we skip the entire 5-layer cascade (YouTube, IZGIV APIs,
    // Cobalt, etc.). Bot-preheat is run when the user types /play, so by
    // the time playback reaches the track, the cache is almost always
    // warm. The resolved URL points at this proxy's own /cache/audio
    // endpoint, which streams from the bot's loopback.
    const botHit = await checkBotCache(cleanId).catch(() => null);
    if (botHit) {
      const proxyAudioUrl = `/cache/audio?yt_id=${encodeURIComponent(cleanId)}`;
      streamCache.set(cleanId, {
        url: proxyAudioUrl,
        expiresAt: Date.now() + CACHE_TTL_MS,
        source: 'bot_cache'
      });
      console.log(`[Resolve] Serving from bot cache: ${proxyAudioUrl}`);
      return res.status(200).json({
        url: proxyAudioUrl,
        title: botHit.title || 'YouTube Audio Feed',
        duration: botHit.duration || 0,
        thumbnail: `https://i.ytimg.com/vi/${cleanId}/hqdefault.jpg`,
        artist: botHit.artist || 'Various Artists',
        format: 'mp3',
        source: 'bot_cache',
        cached: true
      });
    }

    // Resolve stream and metadata in parallel safely
    const [resolved, meta] = await Promise.all([
      masterResolve(cleanId).catch(err => {
        console.error('Error in masterResolve:', err);
        return null;
      }),
      getVideoMetadata(cleanId).catch(err => {
        console.error('Error in getVideoMetadata:', err);
        return {
          title: 'YouTube Audio Feed',
          duration: 0,
          thumbnail: `https://i.ytimg.com/vi/${cleanId}/hqdefault.jpg`,
          artist: 'Various Artists'
        }
      })
    ]);

    if (!resolved || !resolved.url) {
      return res.status(500).json({ error: 'Failed to extract downloadable audio stream from all stages' });
    }

    // Cache the resolved URL!
    streamCache.set(cleanId, {
      url: resolved.url,
      expiresAt: Date.now() + CACHE_TTL_MS,
      source: 'cascade'
    });

    res.status(200).json({
      url: resolved.url,
      title: meta.title,
      duration: meta.duration || resolved.duration || 0,
      thumbnail: meta.thumbnail,
      artist: meta.artist,
      format: 'm4a',
      source: 'cascade'
    });
  } catch (err) {
    console.error('❌ Master resolution failed:', err);
    res.status(500).json({ error: err.message });
  }
});

// PREHEAT — resolve and cache stream URL without proxying the actual audio bytes.
// Called by the CF Worker (or frontend) to warm the cache so that when the
// queue advances, /api/stream serves from cache instantly.
app.get('/api/preheat', async (req, res) => {
  try {
    const videoId = req.query.v || req.query.url;
    if (!videoId) {
      return res.status(400).json({ error: 'Missing parameter "v" or "url"' });
    }

    const cleanId = videoId.startsWith('http') ? ytdl.getVideoID(videoId) : videoId;

    const wasCached = (() => {
      const c = streamCache.get(cleanId);
      return !!(c && Date.now() < c.expiresAt);
    })();

    if (wasCached) {
      console.log(`🔥 Preheat cache HIT for ${cleanId}`);
      return res.status(200).json({ success: true, alreadyCached: true, strategy: 'cache' });
    }

    const resolved = await masterResolve(cleanId).catch(err => {
      console.error(`[Preheat] masterResolve error for ${cleanId}:`, err);
      return null;
    });

    if (!resolved || !resolved.url) {
      console.warn(`[Preheat] All layers failed for ${cleanId}`);
      return res.status(404).json({ success: false, error: 'All resolvers failed' });
    }

    streamCache.set(cleanId, {
      url: resolved.url,
      expiresAt: Date.now() + CACHE_TTL_MS
    });

    console.log(`🔥 Preheated ${cleanId} (cached for ${CACHE_TTL_MS / 60000} min)`);
    res.status(200).json({ success: true, alreadyCached: false, strategy: 'fresh' });
  } catch (err) {
    console.error('❌ Preheat failed:', err);
    res.status(500).json({ success: false, error: err.message });
  }
});

// Stream Audio Direct Proxy
app.get('/api/stream', async (req, res) => {
  try {
    const videoId = req.query.v || req.query.url;
    if (!videoId) {
      return res.status(400).json({ error: 'Missing parameter "v" or "url"' });
    }

    const cleanId = videoId.startsWith('http') ? ytdl.getVideoID(videoId) : videoId;
    console.log(`🎧 Master S-Rank streaming request for: ${cleanId}`);

    // ── LAYER -1 (PRIMARY): Bot-side Telegram-channel cache ──
    // Skip the entire 5-layer cascade on a hit. This proxy exposes its
    // own /cache/audio endpoint that streams from the bot's loopback
    // (see proxyBotAudio above). The redirect target is on the same
    // public host, so the browser stays connected to one origin.
    const botHit = await checkBotCache(cleanId).catch(() => null);
    if (botHit) {
      const audioProxy = `/cache/audio?yt_id=${encodeURIComponent(cleanId)}`;
      console.log(`[Stream] Bot cache HIT → 302 to ${audioProxy}`);
      return res.redirect(302, audioProxy);
    }

    // Fast-path direct redirect to source CDN if requested
    if (req.query.redirect === 'true') {
      console.log(`⚡ Fast-path redirect requested for: ${cleanId}`);
      const cached = streamCache.get(cleanId);
      if (cached && Date.now() < cached.expiresAt) {
        console.log(`🎯 Cache HIT! Fast-redirecting to cached stream URL: ${cached.url}`);
        return res.redirect(302, cached.url);
      }
      const resolved = await masterResolve(cleanId).catch(() => null);
      if (resolved && resolved.url) {
        streamCache.set(cleanId, {
          url: resolved.url,
          expiresAt: Date.now() + CACHE_TTL_MS
        });
        console.log(`✅ Success! Fast-redirecting to resolved stream URL: ${resolved.url}`);
        return res.redirect(302, resolved.url);
      }
    }

    // Handle HEAD requests immediately for pre-flight stability
    if (req.method === 'HEAD') {
      res.setHeader('Content-Type', 'audio/mp4');
      res.setHeader('Accept-Ranges', 'bytes');
      res.setHeader('Cache-Control', 'public, max-age=3600');
      return res.status(200).end();
    }

    const rangeHeader = req.headers.range;

    // --- Stream Cache HIT ---
    const cached = streamCache.get(cleanId);
    if (cached && Date.now() < cached.expiresAt) {
      console.log(`🎯 Cache HIT! Serving cached stream URL inside HF Node Proxy for: ${cleanId}`);
      try {
        const responseHeaders = {
          'Access-Control-Allow-Origin': '*',
          'Access-Control-Allow-Methods': 'GET, OPTIONS',
          'Access-Control-Allow-Headers': 'Range, Content-Type',
          'Access-Control-Expose-Headers': 'Content-Length, Content-Range, Accept-Ranges',
          'Accept-Ranges': 'bytes',
          'Cache-Control': 'public, max-age=3600',
          'Content-Type': 'audio/mp4'
        };

        const fetchHeaders = {
          'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
        };
        if (rangeHeader) {
          fetchHeaders['Range'] = rangeHeader;
        }

        const streamResponse = await fetchWithTimeout(cached.url, {
          headers: fetchHeaders,
          timeout: 15000
        });

        const contentType = streamResponse.headers.get('content-type') || '';
        if (streamResponse.ok && !contentType.toLowerCase().includes('text/html')) {
          // Mirror content headers
          for (const h of ['content-type', 'content-length', 'content-range']) {
            const val = streamResponse.headers.get(h);
            if (val) responseHeaders[h] = val;
          }

          res.writeHead(streamResponse.status, responseHeaders);
          
          const reader = streamResponse.body.getReader();
          const push = async () => {
            while (true) {
              const { done, value } = await reader.read();
              if (done) break;
              res.write(value);
            }
            res.end();
          };

          push().catch(err => {
            console.error('❌ Pipelining error during read on cached stream:', err);
            res.end();
          });

          req.on('close', () => {
            reader.cancel();
          });
          return;
        } else {
          console.warn(`⚠️ Cached stream URL returned non-OK status ${streamResponse.status}. Invalidating cache.`);
          streamCache.delete(cleanId);
        }
      } catch (err) {
        console.error(`❌ Cached stream fetch failed inside HF Node Proxy: ${err.message}. Invalidating cache.`);
        streamCache.delete(cleanId);
      }
    }

    const resolved = await masterResolve(cleanId).catch(err => {
      console.error('Error in masterResolve inside stream:', err);
      return null;
    });

    if (!resolved || !resolved.url) {
      return res.status(500).json({ error: 'Failed to extract audio stream from all stages' });
    }

    // Cache the resolved working URL!
    streamCache.set(cleanId, {
      url: resolved.url,
      expiresAt: Date.now() + CACHE_TTL_MS
    });
    
    // Configure response headers matching cross-origin standards for WebViews
    const responseHeaders = {
      'Access-Control-Allow-Origin': '*',
      'Access-Control-Allow-Methods': 'GET, OPTIONS',
      'Access-Control-Allow-Headers': 'Range, Content-Type',
      'Access-Control-Expose-Headers': 'Content-Length, Content-Range, Accept-Ranges',
      'Accept-Ranges': 'bytes',
      'Cache-Control': 'public, max-age=3600',
      'Content-Type': 'audio/mp4'
    };

    const fetchHeaders = {
      'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
    };

    if (rangeHeader) {
      fetchHeaders['Range'] = rangeHeader;
      console.log(`⚡ Proxying stream with range: ${rangeHeader}`);
    }

    // Proxy the stream response directly from the resolved URL
    const streamResponse = await fetchWithTimeout(resolved.url, {
      headers: fetchHeaders,
      timeout: 15000
    });

    const contentType = streamResponse.headers.get('content-type') || '';
    const ct = contentType.toLowerCase();
    if (!streamResponse.ok || ct.includes('text/html') || ct.includes('application/json') || ct.includes('text/plain')) {
      console.warn(`⚠️ Upstream stream fetch failed or returned invalid content-type ${contentType} (status: ${streamResponse.status}). Falling back to HTTP 302 redirect directly to source CDN!`);
      return res.redirect(302, resolved.url);
    }

    // Mirror content headers
    for (const h of ['content-type', 'content-length', 'content-range']) {
      const val = streamResponse.headers.get(h);
      if (val) responseHeaders[h] = val;
    }

    res.writeHead(streamResponse.status, responseHeaders);
    
    // Convert ReadableStream to Node.js stream and pipe to express response
    const reader = streamResponse.body.getReader();
    const push = async () => {
      while (true) {
        const { done, value } = await reader.read();
        if (done) break;
        res.write(value);
      }
      res.end();
    };

    push().catch(err => {
      console.error('❌ Pipelining error during read:', err);
      res.end();
    });

    req.on('close', () => {
      console.log('🔌 Client disconnected, closing reader');
      reader.cancel();
    });

  } catch (err) {
    console.error('❌ Master streaming error:', err);
    if (!res.headersSent) {
      res.status(500).json({ error: err.message });
    }
  }
});

app.get('/api/test-ytdlp', async (req, res) => {
  try {
    let cmd = req.query.cmd;
    if (cmd) {
      console.log(`[Diagnostic] Executing custom command: ${cmd}`);
      const { stdout, stderr } = await execAsync(cmd).catch(err => ({ stdout: '', stderr: err.message }));
      return res.json({ command: cmd, stdout, stderr });
    }

    const videoId = req.query.v || 'jotdDIy1F90';
    console.log(`[Diagnostic] Running double-engine test for: ${videoId}`);
    
    // 1. Test pytubefix
    const pytubeCmd = `python3 resolve_pytubefix.py "${videoId}"`;
    const pytubeRes = await execAsync(pytubeCmd).catch(err => ({ stdout: '', stderr: err.message }));
    
    // 2. Test yt-dlp
    const videoUrl = `https://www.youtube.com/watch?v=${videoId}`;
    const ytdlpCmd = `yt-dlp -j --format "140/ba/best" --socket-timeout 10 --no-check-certificate --legacy-server-connect --remote-components ejs:github "${videoUrl}"` + 
      (fs.existsSync(COOKIES_PATH) ? ` --cookies "${COOKIES_PATH}"` : '');
    const ytdlpRes = await execAsync(ytdlpCmd).catch(err => ({ stdout: '', stderr: err.message }));

    res.json({
      videoId,
      pytubefix: {
        command: pytubeCmd,
        stdout: pytubeRes.stdout,
        stderr: pytubeRes.stderr
      },
      ytdlp: {
        command: ytdlpCmd,
        stdout: ytdlpRes.stdout.substring(0, 1000),
        stderr: ytdlpRes.stderr.substring(0, 1000)
      }
    });
  } catch (err) {
    res.status(500).json({ error: err.message });
  }
});

app.listen(PORT, '0.0.0.0', () => {
  console.log(`🚀 Node S-Rank YT-DLP / Innertube Proxy listening on port ${PORT}`);
});