bmad-distillator

To compress multiple documents into a single token-efficient output, provide source document paths and an optional token budget. The process generates a lossless distillate that preserves all source facts for LLM-friendly context.

Lossless document distillation is a compression technique that removes overhead while preserving every fact, decision, constraint, and relationship from the source documents. It ensures deterministic, source-preserving compression for downstream workflows.

You can validate your compressed distillate by using the --validate flag. This triggers round-trip checks to ensure fidelity across transformations, verifying that all original source information is retained.

Document compression supports routing, grouping, and optional semantic splitting for large inputs. This allows you to manage multiple source documents efficiently while maintaining deterministic, source-preserving compression.

To generate a token-efficient distillate, you need input paths via source_documents and optional parameters including downstream_consumer, token_budget, output_path, and a --validate flag for round-trip checks.