rcode-memory-distill

To compress a memory bank for faster LLM context loading, generate token-optimized distillates from source files. This process strips presentation overhead while preserving every fact, decision, constraint, and relationship to reduce token consumption during AI context recovery.

Memory bank distillation is a token optimization process that compacts large source files into fast-loading summaries. It preserves lossless facts and relationships while outputting project and stack distillates for rapid AI context restoration without repeated source reads.

To regenerate markdown distillates after memory bank changes, run the distillation process against the updated source files. It uses source digests for change detection, automatically skipping no-op tasks when files remain unchanged to ensure safe, predictable repeated runs.

Yes, token optimization for LLM workflows uses digest-based change detection to skip unchanged source files. This no-op skip mechanism makes repeated distillation runs safe and predictable by avoiding unnecessary processing on unmodified memory bank entries.

Yes, you can distill specific stack information separately from the full project context. The process generates two distinct outputs: a comprehensive project distillate and a stack-only distillate, both written to the distillates directory.