upstash/semantic-cache

Leverages Upstash Vector databases with configurable embedding models to store and retrieve cache entries by semantic similarity score rather than exact string matching. Includes namespace partitioning for multi-tenant isolation and adjustable `minProximity` thresholds (0-1) to control match sensitivity, enabling synonym recognition and multilingual query support across Node.js applications. Primary use case is reducing redundant LLM API calls by matching semantically similar queries to cached responses.

293 stars and 5,659 monthly downloads. No commits in the last 6 months. Available on npm.