mindtro/semafold

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).

33
/ 100
Emerging

This project helps AI infrastructure teams reduce the storage footprint of numerical AI data. It takes large embedding vectors or key-value tensors and outputs significantly smaller, compressed versions, maintaining data quality. This is for engineers and architects building AI systems who need to manage large volumes of AI-specific data more efficiently.

Use this if you need to dramatically shrink the size of your AI embeddings or KV-cache tensors to save storage and improve performance within your AI infrastructure.

Not ideal if you're looking to summarize text or reduce the number of tokens in natural language processing tasks, as it focuses purely on numerical data compression.

AI-infrastructure vector-databases machine-learning-operations embedding-storage inference-optimization
No Package No Dependents
Maintenance 13 / 25
Adoption 6 / 25
Maturity 9 / 25
Community 5 / 25

How are scores calculated?

Stars

17

Forks

1

Language

Python

License

Apache-2.0

Category

database

Last pushed

Apr 01, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/vector-db/mindtro/semafold"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.