deploy

name	deploy
description	Publish the current honest metrics to README, website and app (this project's only "deploy").

There is no server deployment. "Deploy" means publishing the latest honest numbers from the single source of truth:

1. Fresh artifacts first: python -m src.models.train then python -m src.models.backtest (they write models/atp_metrics.json and reports/last_backtest.json).
2. python scripts/publish_metrics.py — rewrites docs/web/metrics.js and the README <!--METRICS--> block from those files.
3. python scripts/log_metrics_history.py — appends the snapshot to reports/metrics_history.csv.
4. Commit with a docs or web scope.

Rules: never publish numbers that didn't come from a retrain you actually ran — a merged experiment is not a verified one (EXPERIMENTS.md, 2026-06-12 lesson). The website is GitHub Pages off this repo; the desktop app reads atp_metrics.json live via /api/model, so it needs no extra publish step.