How would you implement caching for a GraphQL API, given that unlike REST, a single endpoint serves widely varying query shapes that don't map cleanly to standard HTTP caching?

Because GraphQL typically uses a single POST endpoint (making standard HTTP/CDN caching based on URL largely ineffective, since every distinct query looks identical at the HTTP level) and different clients request different field combinations, caching strategies instead operate at other levels: persisted queries (registering common queries server-side ahead of time and referencing them by a short hash, which also enables safe GET-based caching), response caching keyed by the specific query and variables combination, and field-level or entity-level caching (like Apollo Server's cache hints, specifying how long a specific type or field's data should be cached) integrated with a backing store like Redis.

const resolvers = {
  Query: {
    product: (_, { id }, { cache }) => {
      // Apollo cache hints: this field's result can be cached for 60 seconds
      return cache.wrap(`product:${id}`, () => db.products.findById(id), { ttl: 60 });
    },
  },
};