Vector compression with TurboQuant codecs for embeddings, retrieval, and KV-cache. 10x compression, pure NumPy core — optional GPU acceleration via PyTorch (CUDA/MPS) or MLX (Metal).
QJL sign-based vector compression and scoring in Rust — near-optimal distortion rate, append-only persistence, CPU-only, no LLM
Source code for ICML'26 paper "RQ-MoE: Residual Quantization via Mixture of Experts for Efficient Input-Dependent Vector Compression".
ChatMind: Semantic search for Discord & KakaoTalk chat messages. Search by meaning, not keywords. Powered by TurboQuant compression (ICLR 2026).
CommitMind: Semantic search for Git commit history powered by TurboQuant vector compression (ICLR 2026). Search commits by meaning, not just keywords.
AI-powered log anomaly detection CLI — learns normal patterns, detects anomalies with semantic embeddings, matches past incidents. Powered by TurboQuant 3-bit compression (ICLR 2026).
From-scratch survey of vector compression (JL, SVD, sign/scalar quantization, TurboQuant) evaluated on RAG retrieval over a linear-algebra textbook
Accelerate LLM KV cache compression with a PyTorch TurboQuant implementation for efficient, high-quality vector quantization.
To associate your repository with the vector-compression topic, visit your repo's landing page and select "manage topics."