thom-heinrich/chonkify

Extractive document compression for RAG and agent pipelines. +69% vs LLMLingua, +175% vs LLMLingua2 on information recovery. Compiled wheels, try it out.

39
/ 100
Emerging

Builds document units scored through 768-dimensional embeddings and selects the highest-ranked segments to stay within token budgets while maximizing factual recovery—critical for quantitative research and reasoning traces where exact facts outweigh fluent paraphrasing. Supports multiple embedding backends including Azure OpenAI, OpenAI-compatible APIs, and fully offline local SentenceTransformers, with a CLI and Python API for RAG pipelines and agent memory systems. Ships as compiled extension modules for performance-sensitive workloads across Linux, Windows, and macOS platforms.

No Package No Dependents
Maintenance 13 / 25
Adoption 4 / 25
Maturity 9 / 25
Community 13 / 25

How are scores calculated?

Stars

5

Forks

2

Language

Python

License

Last pushed

Mar 26, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/rag/thom-heinrich/chonkify"

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