Pivot ecomm from a generic multi-tenant Shopify-alternative storefront to the
commitment-commerce + cross-maker verified referral NETWORK direction
(ecomm-content/rfcs/network_strategy.md; OHM identity/relationality super-drafts).
Phase A of the reset — clear the decks:
- Strip the storefront surfaces with no carry-forward: the storefronts domain,
the products import/export/image HTTP spine (service/imagefetch/repo + endpoints),
the SPA frontend, and the Playwright e2e suite.
- Keep the reusable core: /healthz + /api/auth/* (accounts); the
catalog-normalization library (codec/dialect/models/serialize/validate/diff/hosted —
importable, no HTTP yet); platform/* infra. Migrations untouched (append-only).
- Reduce check.sh to backend-only (import-linter + pytest); trim dev.sh and the unused
GCS overlay env. Repoint app.json/README/CLAUDE.md; bump VERSION 0.8.0 -> 0.9.0.
check.sh green: import-linter (main > domains > platform) KEPT, pytest passing.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two-step email + one-time-code sign-in behind both doors; pure entry-routing rule unit-
tested with Vitest (§6.8); storefront routing reaches a SLICE-2 signed-in placeholder with
sign-out (PUC-9). check.sh now runs the frontend unit tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
macOS ships bash 3.2 (no wait -n): poll with kill -0 instead. npm spawns vite as a
grandchild, so cleanup now recurses pgrep -P to reap the whole subtree — Ctrl-C frees
both :8000 and :5173 cleanly. Verified clean-checkout -> dev.sh -> empty self-migrated
Postgres -> /healthz 200, and trap teardown.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ben.stull