VQF-1 compression

VQF-1 is a draft physical compression for V5 object and event segments. KIP-0027 is still a draft. The format is not frozen.

knolo v5 compress accepts the profiles fast, balanced, and max. knolo v5 decompress restores ordinary required segments. Run knolo v5 compress --help for the flags in the CLI version you installed.

Compression preserves logical roots. A compressed object or event image needs a VQF-aware reader. v5 decompress restores ordinary required segments.

TypeScript is the encoder. Rust and Python can decode required object and event segments and check the same logical payload, identity, segment root, and state root. Native VQF encoding, lexical postings, and the optional query-index codec stay outside the portable Rust writer. Lexical postings remain a TypeScript runtime artifact.

Do not treat a VQF size number from an internal benchmark note as a guarantee for your corpus. The repository keeps those measurements in docs/VQF1_BENCHMARKS.md. They are not a product SLO.

Until the format freezes, keep an uncompressed image when another runtime must mount the bytes without a VQF reader.