Skip to content

README.md

Latest commit

 

History

History
40 lines (33 loc) · 2.33 KB

README.md

File metadata and controls

40 lines (33 loc) · 2.33 KB
Raw

levels/

Level data for the game, at three stages of a pipeline. Most of what lands here is transient and git-ignored — only the curated ladders are tracked, because they're the durable, reviewable output.

Lifecycle

campaign / gen ──▶ candidate pool ──▶ curate ──▶ ladder ──▶ wire ──▶ game
  (search)          gen-*.json /        (pick)    ladder*.json  (place    src/game/
                    campaigns/<name>/             (tracked)     info +    levels.json
                    pool.json (ignored)                         résumé)   (tracked)
  1. Search — level-gen campaign --name <run> (or gen --batch --profile spread) generates and screens millions of boards. Output:
    • campaigns/<name>/pool.json + manifest.json — a resumable top-K pool. Ignored.
    • gen-*.json — a flat candidate pool from gen --batch. Ignored.
    • report.md — a browsable per-bucket ranking (level-gen rank). Ignored.
  2. Curate — level-gen curate --from-campaign <run> --profile initial128 picks a teaching-ordered ladder and writes ladder128.json (and ladder32.json for the shorter profile). Tracked — these are the reproducible, reviewed end product.
  3. Wire — python3 level-gen/wire_ladder.py merges ladder128.json with the résumé content from assets/resume.json and writes src/game/levels.json (placing one info tile on the first 16 levels). Tracked.

What's tracked vs. ignored

Path Tracked? Why
ladder32.json ✅ Curated short ladder (reproducible end product)
ladder128.json ✅ Curated full ladder (wired into the game)
gen-*.json ❌ Transient candidate pool; reproducible by seed
campaigns/<name>/ ❌ Transient campaign state; resumable by seed
report.md ❌ Generated ranking report

The ignored artifacts are large and fully reproducible from the seed range, so they stay out of git. Regenerating a ladder is a deliberate, reviewed step — see ../level-gen/README.md.