Cachet – A drop-in semantic cache for LLM APIs, 100% local, in Rust
Semantic caching without a vector DB—just swap your base URL.
Vector search on SQLite alone hits 21ms p.95 — no vector database required.
ML engineers, developers building RAG applications
Chroma · LanceDB · Qdrant
Semantic caching without a vector DB—just swap your base URL.
Using a single-file .pardus format with CREATE/INSERT/SELECT + SIMILARITY queries gives a very familiar developer UX for embedding storage. The combination of graph-based ANN, full transactions, thread-safety, and zero external dependencies is an uncommon and useful engineering combo for local-first AI work; it would win more attention with benchmark comparisons and richer ecosystem integrations (connectors/clients).
Self-managed vector DB in Rust with <5ms queries; Pinecone/Weaviate already own this space.
Replaces Windows Search with vector + full-text hybrid; EXIF geocoding and function-level code chunking actually work.
Pandoc reimplementation in Rust with 16x speedup but incomplete format support.
Local semantic search beats Everything and ripgrep; adds OCR, GPS reversal, MCP agents.