Resilience Core Skill

This skill ensures 100% uptime for metadata fetching by using a multi-layer fallback strategy. Metric: The user must NEVER see a blank poster or missing description due to a network error.

🌐 The Multi-Level Resilience Cascade

Every external request (especially to TMDB) MUST go through tmdbClient.js. NEVER use fetch() directly for metadata.

Cascade Order:

1. Custom Cloudflare Worker: First line of defense.
2. Lampa Proxy: Public mirror (apn-latest.onrender.com).
3. Server Proxy: Self-hosted proxy (/api/proxy?url=...).
4. CapacitorHttp + Client DoH (Native Only):
   • Uses dns.google API to resolve IP, bypassing ISP DNS Poisoning.
   • Sends direct HTTPS requests to IP with Host header.
5. Corsproxy.io: Browser-based fallback.
6. Kinopoisk (Out-of-band Fallback):
   • Used ONLY for text data (titles, descriptions).
   • Triggered ONLY if TMDB is completely unreachable via all above levels.

🚦 Traffic Isolation Rule

RULE: DoH (DNS-over-HTTPS) and IP-direct requests are used TOKYO for API data (JSON).

• NEVER use DoH mechanisms for loading images/posters.
• Images have their own resilience logic (Mirrors -> WSRV.NL).

🛡️ Image Resilience

Images use a separate logic:

• Mirrors: imagetmdb.com, nl.imagetmdb.com, etc.
• Auto-Ban: If a mirror fails 20 times in 10s, it’s banned.
• WSRV.NL: If all mirrors fail, we switch to wsrv.nl proxying.

💻 Usage Example

import tmdbClient from '../utils/tmdbClient';

// BAD ❌
// const res = await fetch('https://api.themoviedb.org/3/movie/550');

// GOOD ✅
const data = await tmdbClient('/movie/550');
if (data.source === 'kinopoisk') {
    // Handle specific KP logic if needed
}