redis-core

Redis Core is a data-modeling guidance skill from Redis, Inc. that helps developers make the two decisions which most affect Redis memory, performance, and maintainability: which data structure to use, and how to name keys. It solves the recurring problem of applications defaulting to a serialized string blob for everything, then paying fetch-parse-mutate-rewrite costs on every field update.

The skill centers on a decision matrix that maps access patterns to Redis types: Strings for simple values and atomic counters, Hashes for objects with independently updated fields, Lists for queues and recent-N items, Sets for uniqueness and membership, Sorted Sets for rankings and leaderboards, JSON for nested/hierarchical data with path-level updates, Streams for event logs and fan-out messaging, and Vector Sets for vector similarity with HNSW. It calls out the flat-object-in-a-string anti-pattern and recommends Hashes instead. The second half establishes key-naming conventions: lowercase colon-separated segments in a stable hierarchy (for example user:1001:profile), short but readable keys, tenant prefixes for multi-tenancy, and avoiding full URLs as keys. Reference files provide fuller rationale plus Python (redis-py) and Java (Jedis) examples, and the package ships eval suites and benchmark baselines used to measure the skill's effect across models.

The intended users are backend and platform engineers designing a Redis data model, building caches, counters, leaderboards, membership sets, or session stores, or reviewing and refactoring existing key names. It is a read-only conceptual reference: it prescribes conventions and shows illustrative code but does not execute anything against a live Redis instance, making it a safe design companion.