bmad-distillator

Document distillation compresses multiple source documents into a single lossless distillate optimized for LLM context, preserving facts, decisions, constraints, and relationships for downstream workflows.

Lossless document distillation compresses source documents into an LLM-optimized distillate through analysis, compression, and verification stages, ensuring facts, decisions, and relationships are preserved without data loss.

Run round-trip validation with the --validate flag to verify losslessness after document distillation, checking completeness and format to ensure all critical information from source documents is preserved.

Yes, distillation can route, group, and semantically split source documents into multiple distillates for downstream workflows, producing semantically segmented outputs without reading the actual content.

Specifying a token budget is optional for document distillation, allowing you to optimize the final distillate size to fit the context constraints of your specified downstream LLM consumer.

Semantic splitting and routing occur without reading content, meaning losslessness depends on the verification stage, which should be run with --validate to confirm completeness for downstream workflows.