bmad-distillator

Document distillation compresses multi-file source documents into a token-efficient distillate without losing facts. It preserves named entities, constraints, and relationships while deduplicating overlap to provide dense downstream context for LLM workflows.

Lossless compression reduces tokens in research packs by removing redundancy and filler while preserving scope boundaries, risks, and rationale. It organizes themes into dense, thematically grouped bullets to provide dense downstream context for PRD creation.

Yes, you can distill large multi-file documentation sets using semantic splitting. When a single distillate is too large, semantic splitting divides the compressed output into manageable segments while preserving facts, decisions, and relationships for downstream workflows.

Yes, document distillation includes completeness checks and round-trip validation for high-stakes use cases. This verification ensures lossless compression faithfully preserves all facts, decisions, and constraints from the original source documents without losing important relationships.

Lossless document compression preserves facts, decisions, constraints, named entities, headings, numbers, open questions, scope boundaries, risks, and rationale. It removes redundancy and filler while maintaining relationships between these elements for downstream analysis.