Module codec 

Codec orchestrator — compress and decompress through the DataCortex pipeline.

Phase 1: Format preprocessing + zstd (Fast mode). Phase 3: Full CM engine with higher-order models + mixer + APM (Balanced mode, ~256MB). Phase 5: Full CM engine with 2x context maps (Max mode, ~512MB). Phase 6: Dual-path CM + LLM with MetaMixer (Max mode with neural feature). Phase 7: Dual-path CM + GRU byte-level predictor (Balanced mode).

Functions

compress

Compress data into .dcx format, writing to output.

compress_to_vec

Compress to Vec (convenience).

compress_to_vec_with_model

Compress to Vec with explicit model path.

compress_to_vec_with_options

Compress to Vec with explicit model path and zstd level override.

compress_with_full_options

Compress with all options including external dictionary.

compress_with_model

Compress with optional explicit model path (for neural Max mode).

compress_with_options

Compress with optional explicit model path and zstd level override.

decompress

Decompress a .dcx file from input, returning the original data.

decompress_from_slice

Decompress from slice (convenience).

decompress_with_model

Decompress with optional explicit model path (for neural Max mode).

raw_zstd_compress

Compress raw data with zstd at a given level (for benchmark comparison).

read_header

Read header only (for info command).

train_dict

Train a zstd dictionary from multiple sample files.