Engine documentation

CANON/ : the distilled core

Engine protocol and working reference.

Generated by python3 generate_canon.py (run from the package root):

  • KERN.md : the core context injected at every session start. Generated section between the GENERATED markers; your own text between <!-- HAND --> and <!-- /HAND --> markers survives every re-run.
  • LLM-PROMPT.md : a model-agnostic distillation prompt plus a corpus digest. Feed it to any LLM, save the answer, import it with python3 generate_canon.py --from-file <answer>.

Empty until you ingest data and run the generator.