# embedcache

> A Rust library and REST API that generates text embeddings locally using FastEmbed and caches them in SQLite. No external embedding API calls, no per-token billing, no rate limits.

EmbedCache is a project by Skelf Research. It bundles 22+ local embedding models (BGE, MiniLM, Nomic, multilingual E5) via the FastEmbed crate and persists results in a SQLite-backed cache keyed by content hash plus model identifier.

It is distinct from embedding-API caching wrappers like LangChain CacheBackedEmbeddings — EmbedCache replaces the hosted embedder rather than memoizing it.

License: GPL-3.0. Crate: https://crates.io/crates/embedcache. Docs: https://docs.skelfresearch.com/embedcache/.

## Site

- [Home](https://embedcache.skelfresearch.com/): Overview, features, what it is and isn't.
- [About](https://embedcache.skelfresearch.com/about): Why the project exists and who it's for.
- [Notes](https://embedcache.skelfresearch.com/blog): Working notes on embedding caching, cost, and invalidation.
- [Compare](https://embedcache.skelfresearch.com/compare): Honest comparisons with adjacent tools.

## Notes

- [Embedding API spend is a tax on stable inputs](https://embedcache.skelfresearch.com/blog/embedding-spend-is-a-tax-on-stable-inputs/)
- [Content-hash keying vs LRU: which actually saves money](https://embedcache.skelfresearch.com/blog/content-hash-vs-lru/)
- [Stale-while-revalidate for embeddings: the corner cases](https://embedcache.skelfresearch.com/blog/stale-while-revalidate-for-embeddings/)

## Compare

- [EmbedCache vs LangChain CacheBackedEmbeddings](https://embedcache.skelfresearch.com/compare/langchain-cachebackedembeddings/)
- [EmbedCache vs raw Redis as embedding cache](https://embedcache.skelfresearch.com/compare/redis-as-embedding-cache/)

## External

- [Documentation](https://docs.skelfresearch.com/embedcache/)
- [GitHub](https://github.com/skelfresearch/embedcache)
- [Crates.io](https://crates.io/crates/embedcache)