bmad-distillator

You can distill documents by asking the AI to analyze your source files and route them between single-pass or fan-out semantic compression. It outputs a high-signal context package, verifying completeness against headings and named entities to ensure no facts are lost.

The best way to prepare multiple architecture documents for PRD drafting is through lossless distillation. It applies thematic deduplication and completeness verification to convert your file set into a single, dense context package optimized for downstream language model workflows.

Yes, you can validate information retention using optional round-trip reconstruction. The process checks completeness against headings and named entities, measures the compression ratio, and reconstructs source documents to confirm that facts and relationships are accurately preserved.

Yes, document distillation works with large sets of discovery notes and research reports by applying fan-out compression and semantic splitting. It routes large outputs into self-contained topical sections, applying thematic deduplication to ensure token-efficient context packaging.

Use semantic splitting instead of single-file compression when processing large document sets that exceed optimal token limits. The routing mechanism analyzes your files and splits outputs into self-contained topical sections to maintain complete, lossless context for downstream LLM workflows.