5 projects
sentrygate
A local layer that catches prompt injections and masks private data out of a prompt before it leaves your machine. On-device redaction plus a fast 421M decision-model guard, no second LLM and no GPU.
turboqvec
Compressed vector store using PolarQuant — 8-15x smaller float arrays with fast inner product queries
pageindex-rs
Hierarchical document indexing and tree-based retrieval for RAG pipelines, written in Rust
fastrustrag
FastRustRAG: Blazing-fast document deduplication using MinHash and LSH - 50-100x faster than Python
aqiretrieve
A python package for retrieving AQI from EPA
