The project compresses JSON results from AI agents by minifying, tabularizing, and dictionary‑coding data, keeping a lossless round‑trip guarantee. Users can define per‑tool policies to optionally apply lossy reductions on chosen fields. It targets developers building AI pipelines who need to stay within token limits without sacrificing data fidelity. Compared to generic minifiers, it offers structured compression and policy‑driven control tailored for LLM contexts.
View on GitHub →inth3shadows/terse