What it solves
Trainable text compression benefits from frequency-aware coding rather than a fixed generic codec.
How it works
1Training corpus→2codec model→3encode/decode API→4v2 or v3 stream
Use it when
Train from representative text
Use v2 or matrix-based v3 streams
Choose a documented deployment objective
Quick Start
#include "loxc.h"
/* Train or load a documented model, then encode/decode via the public API. */Engineering evidence
- README documents trainer, CLI, simple API, examples, tests, benchmarks and technical documentation
- v3 documents HIER4/HIER8 nested matrices and diagonal cells that save one coordinate value
Known limits
- Compression trade-offs depend on the selected corpus and objective
- Benchmark interpretation requires the documented corpus and configuration