# Embedding cache (~6MB of vectors) — a pure speed/cost optimization, fully
# regenerated by rerunning sweep.py. Never a source of truth.
cache/
__pycache__/
