Files
bitgate/examples/commerce/README.md
T
thePR0M3TH3AN 2beb7a7bb5 feat: adapter contract, commerce proof-of-fit, and reusability docs
Proves the engine did not stay video-specific during extraction.

- application adapter contract in core (createApplicationAdapter,
  evaluateObject, collectTargets): an object names the targets that govern it
  and the strictest verdict reaching any of them wins, so denying a seller
  takes their whole catalogue down with no cascade logic in the application
- examples/commerce: a marketplace on the public API covering all ten
  proof-of-fit behaviors the plan requires, including that no core change was
  needed. One product, one snapshot, four surfaces, four different verdicts.
- README, integration guide, per-example README, CHANGELOG, four ADRs, and
  GPL-3.0-or-later LICENSE (the license was listed in Phase 1 but missing)
- checkpoints rewritten to reflect actual state and to say plainly which
  migration-plan phases belong to the consumer repos rather than here

Fixes a real gap found by the proof-of-fit: composeDecisions kept only the
first decision's evidence, so an object governed by several targets lost the
evidence that drove its verdict. Evidence now merges — counts summed,
contributor lists unioned, flags or-ed.

556 tests; lint, typecheck, and per-package type build clean.
2026-08-09 17:57:09 -04:00

69 lines
2.9 KiB
Markdown

# Commerce proof-of-fit
A marketplace built on `nostr-governance`, using only the published API.
The point is not to ship marketplace moderation. It is to prove the engine did
not accidentally stay video-specific during extraction: if a commerce consumer
needs a change to `@nostr-governance/core`, the abstraction is wrong.
## What it demonstrates
The migration plan lists ten behaviors a commerce consumer should show before
`1.0.0`. Each is a numbered block in
[`test/proof-of-fit.test.js`](test/proof-of-fit.test.js):
1. **Seller allowlist mode** — unlisted sellers are hidden from discovery, but
the allowlist is not enforced at checkout
2. **Administrative seller denial** — takes the seller's whole catalogue down
while leaving them their own dashboard view
3. **Product-address denial** — survives the seller republishing under a new
event id, and does not touch their other listings
4. **Exact-event denial** — affects only the named revision
5. **Trusted scam-report downranking** — one report reorders, three block
checkout; nothing is hidden on a single signal
6. **Malware checkout denial** — one trusted malware report blocks the sale,
while three "misleading" reports do not
7. **Moderator capability checks** — a listing moderator cannot deny a seller;
a seller moderator cannot deny a listing address
8. **Community curator ingestion** — curated denials are honored, curator
contributions beyond their capability are dropped, and the root cannot be
denied through a curated list
9. **Seller-dashboard explanation** — the seller sees who reported and why;
public surfaces see only counts
10. **No core changes required** — four different verdicts for one product from
one snapshot, using only exported functions
## Files
| File | Contents |
| --- | --- |
| `src/policy.js` | Marketplace thresholds and roles. None of these numbers exist in the core. |
| `src/adapters.js` | Products, storefronts, reviews, and sellers as governance targets. |
| `src/marketplace.js` | Feed pipeline, grid ordering, checkout, seller explanation. |
## The shape worth copying
A product answers to three targets — its seller, its exact event, and its
address coordinate — and the composed decision is the strictest verdict reaching
it. That is what makes "deny the seller" remove every listing without a line of
cascade logic in the application.
One product, one snapshot, four surfaces:
| Surface | Visibility | Transaction |
| --- | --- | --- |
| `public-marketplace` | `restrict` | `deny` |
| `product-detail` | `restrict`, overridable | `deny` |
| `checkout` | `allow` | `deny` |
| `seller-dashboard` | `restrict`, with evidence | `deny` |
Reviews deliberately answer to their reviewer and their own event — never to the
product they review. Otherwise denying a bad product would silence the reviews
warning people about it.
## Run it
```bash
npm test -- examples/commerce
```