bmad-distillator

You compress source documents into a lossless distillate by stripping prose, hedging, and formatting overhead while preserving every fact, decision, constraint, and relationship. This minimizes token usage for downstream LLM workflows.

Lossless distillation is the process of compressing documents into a dense format that retains all core facts and relationships. It removes human-centric overhead to produce hyper-compressed, LLM-ready context for knowledge consolidation.

You generate a distillate by providing source documents and an optional downstream consumer. The process applies stage-driven processing and semantic splitting to output either a single dense distillate or semantically segmented sections.

Yes, the distillation process supports semantic splitting to produce semantically segmented section distillates. This allows you to organize hyper-compressed context into specific segments that match downstream LLM pipeline requirements.

The best way to reduce token usage is to compress large document sets into a lossless distillate. This strips formatting and prose while strictly preserving facts and relationships, ensuring token efficiency for downstream LLM workflows.

Yes, the distillation process includes optional validation alongside strict distillate-format rules. This ensures the generated output satisfies frontmatter metadata requirements and preserves structural integrity for downstream LLM pipelines.