Show HN: Valkey-native semantic cache with OTel and Prometheus built in

4 points

2 months ago

Every semantic cache library - RedisVL, LangChain, LiteLLM - targets Redis and assumes RediSearch wire compatibility. Point any of them at Valkey Search 1.2 and they break silently in multiple places. There's no library that actually handles Valkey natively. This does.

@betterdb/semantic-cache is a standalone semantic cache backed by valkey-search. No framework coupling - works with any LLM client. Every check() and store() emits an OTel span and increments Prometheus counters and histograms out of the box. No instrumentation code required. MIT. Optional LangChain and Vercel AI SDK adapters. Works on self-hosted Valkey, ElastiCache for Valkey, and Memorystore for Valkey.

https://www.npmjs.com/package/@betterdb/semantic-cache

The Valkey Search 1.2 divergences from RediSearch that required explicit handling are in the changelog.

1 comment