bmad-distillator

Lossless document distillation compresses multiple source documents into a single, token-efficient distillate that preserves all facts, decisions, constraints, and relationships for downstream LLM contexts.

Lossless distillation is a compression technique that consolidates multiple input documents into a compact, token-efficient format while strictly preserving all original facts, decisions, constraints, and relationships.

Provide source_documents, downstream_consumer, token_budget, and output_path to distill inputs into a portable distillate with frontmatter and sectioned content, optionally enabling validation with the --validate flag.

Yes, lossless distillation supports semantic splitting into sectional distillates complete with orientation and cross-references, ensuring structured navigation for downstream LLM consumers.

Yes, lossless distillation is specifically useful for PRD-like workflows, architecture reviews, and knowledge baselining across multiple documents while maintaining a strict token budget for LLMs.