bmad-distillator

Document distillation compresses multiple source files into a single token-efficient distillate. This process removes prose overhead and redundancy while preserving every fact, decision, constraint, and relationship for downstream LLM consumption.

Lossless document distillation analyzes and compresses sources into a token-efficient format. It ensures every critical metadata, fact, and relationship is preserved through a stage-wise process of analysis, compression, and verification.

Yes, you can process mixed-source scenarios including single files, folders, or glob patterns. The distillation routine validates these inputs and optionally routes and splits the output based on your specified token budget.

You can set a token budget using the optional token_budget parameter. This drives the splitting and validation stages, ensuring the final lossless distillate fits within your downstream LLM context constraints.

Yes, the process includes a verification stage and an optional round-trip validation step. This ensures the generated distillate losslessly preserves every fact and constraint from the original source documents.