Zero-Cost YouTube Caching on Cloudflare Workers

The Problem with YouTube RSS

The old site used RSS to pull YouTube videos. It looked like this:

const feed = await fetch(
  `https://www.youtube.com/feeds/videos.xml?channel_id=${channelId}`
);

Three problems:

1. YouTube quietly rate-limits or blocks RSS scrapers with no error — videos just disappear
2. RSS doesn’t support playlist-based fetching — you get the full channel feed with no way to filter by playlist
3. XML parsing is CPU-intensive in a Cloudflare Worker (though manageable — see the analysis post)

The Solution: Two-Layer KV Cache

The replacement architecture uses the official YouTube Data API v3 with Cloudflare KV as a cache layer between your Worker and the API.

Layer 1: Channel playlists (auto-discovery, 6hr cache)
─────────────────────────────────────────────────────
KV.get('channel-playlists:{channelId}')
  HIT  → return cached JSON              // ~0.3ms CPU, 99.9% of requests
  MISS → fetch /youtube/v3/playlists     // 1 API unit, ~100-300ms wall-clock
       → KV.put(result, ttl: 6hrs)

Layer 2: Videos per playlist (1hr cache)
─────────────────────────────────────────
KV.get('playlist:{playlistId}')
  HIT  → return cached JSON              // ~0.3ms CPU, 99.9% of requests
  MISS → fetch /youtube/v3/playlistItems // 1 API unit, ~100-300ms wall-clock
       → KV.put(result, ttl: 1hr)

The key insight: I/O time (network fetches, KV reads) doesn’t count toward Cloudflare’s 10ms CPU limit. Only JavaScript execution counts. So a KV cache hit costs ~0.3ms CPU — trivially safe on the free tier.

The Code

export async function getAllPlaylists(
  channelId: string,
  kv: KVNamespace,
  apiKey: string,
): Promise<Playlist[]> {
  const cacheKey = `channel-playlists:${channelId}`;

  // Try cache first
  const cached = await kv.get(cacheKey, 'json') as Playlist[] | null;
  if (cached) return cached; // HIT

  // Cache miss — fetch from API
  const url = new URL('https://www.googleapis.com/youtube/v3/playlists');
  url.searchParams.set('part', 'snippet,contentDetails');
  url.searchParams.set('channelId', channelId);
  url.searchParams.set('maxResults', '50');
  url.searchParams.set('key', apiKey);

  const res = await fetch(url.toString());
  const data = await res.json();
  const playlists = data.items.map(/* ... transform ... */);

  // Store with 6hr TTL
  await kv.put(cacheKey, JSON.stringify(playlists), { expirationTtl: 21600 });
  return playlists;
}

The Quota Math

At 10 playlists, refreshed every hour:

You’d need 41 playlists refreshing every hour to hit the free quota limit. Very safe.

When You’d Need to Pay

The free tier gives you 100K Worker requests/day and 100K KV reads/day. You hit these limits at the same time:

~3,334 daily visitors to dynamic pages = $5/month upgrade trigger

At that point you’re probably monetising anyway. The correct framing: the day you need to pay $5 is the day you have 100K people visiting your site. Not a bad problem to have.