bmad-distillator

You can compress documents losslessly by extracting every discrete fact, decision, constraint, and relationship into a single dense distillate or semantically split sections, minimizing token overhead while preserving original information fidelity.

Lossless distillation is the process of identifying and extracting all discrete facts and relationships from source documents to produce a hyper-compact distillate that preserves original information fidelity for downstream AI tasks.

Yes, the distillation process supports semantically split distillates, allowing you to load specific sections as context to fit varying downstream LLM context window requirements.

You validate distillates using annotated frontmatter that includes source origins, downstream consumer information, and token estimates, enabling round-trip validation and full traceability of the extracted information.

Document distillation applies to a range of source documents including briefs, discovery notes, reports, and architecture docs, extracting facts and relationships to minimize token overhead.

No external dependencies are required to distill documents; you simply provide source documents and an optional downstream consumer parameter to generate the lossless distillate output.