golang-samber-hot

This skill provides guidance for using samber/hot, a type-safe in-memory caching library for Go designed to reduce latency and backend load in applications that repeatedly access the same medium-to-low cardinality data. It focuses on helping developers adopt caching as a deliberate system design decision rather than an implementation detail, emphasizing correct eviction strategy selection and operational awareness. The library supports Go 1.22+ and is intended for high-frequency workloads where repeated reads can be served from memory instead of hitting slower backend systems.

It offers a broad set of eviction algorithms including W-TinyLFU, LRU, LFU, TinyLFU, S3FIFO, ARC, TwoQueue, SIEVE, and FIFO, allowing optimization based on workload characteristics such as recency, frequency, or scan resistance. Additional capabilities include TTL-based expiration, cache loaders with singleflight-style deduplication for missing keys, cache sharding for scalability, stale-while-revalidate behavior to reduce cache stampedes, missing key caching to prevent repeated backend misses, and Prometheus metrics for observability. These features collectively support building robust caching layers that can be tuned for both performance and reliability.

Typical use cases include session caching, frequently accessed user or product data, DNS or lookup-style frequency-heavy datasets, and any service where backend calls are expensive or rate-limited. It is especially relevant for backend Go services handling high request throughput. The intended users are Go engineers and system designers who need fine-grained control over cache behavior and want to improve system responsiveness while reducing infrastructure pressure through intelligent in-memory caching strategies.