docs(readme): revise backend overview, add ingest/payment API docs and seed data details

README.md

@@ -1,6 +1,6 @@
 # monk-code-screen
 
-Billing/subscriptions/invoicing backend skeleton. Bun + Hono + `bun:sqlite` (zero network deps at runtime; TS runs natively, no build step).
+Billing/invoicing/payments backend skeleton. Bun + Hono + `bun:sqlite` (zero network deps at runtime; TS runs natively, no build step).
 
 ## ⚡ START COMMAND (sticky)
 
@@ -20,8 +20,10 @@ bunx tsc --noEmit    # typecheck
 
 ```
 src/index.ts    Hono app + routes
-src/db.ts       opens data.db, applies schema, seeds if empty (runs on import)
-src/schema.sql  customers / plans / subscriptions / invoices (idempotent)
+src/ingest.ts   CSV parser/normalizer + transactional insert (shared by seed and POST /ingest)
+src/db.ts       opens data.db, applies schema, seeds from invoices.csv if empty (runs on import)
+src/schema.sql  customers / invoices / payments (idempotent)
+invoices.csv    seed data — deliberately messy; DO NOT MODIFY
 ```
 
 DB is `data.db` at repo root (WAL, FK on). Boot always re-applies schema, so adding a table = edit schema.sql + restart.
@@ -31,19 +33,29 @@ DB is `data.db` at repo root (WAL, FK on). Boot always re-applies schema, so add
 ```sh
 curl localhost:3000/health
 curl localhost:3000/customers
-curl localhost:3000/customers/cus_ada/invoices
-curl -X POST localhost:3000/subscriptions \
+curl localhost:3000/customers/<cus_id>/invoices
+curl localhost:3000/invoice/INV-1011                          # by invoice_number or surrogate inv_ id
+curl -X POST localhost:3000/ingest --data-binary @invoices.csv
+curl -X POST localhost:3000/payment/INV-1011 \
   -H 'content-type: application/json' \
-  -d '{"customer_id":"cus_grace","plan_id":"plan_basic"}'
+  -d "{\"idempotency_key\":\"$(uuidgen)\",\"amount\":1500000,\"currency\":\"USD\"}"
 ```
 
-`POST /subscriptions` is the happy path: validates with zod, creates subscription + first invoice in one transaction, returns both (201).
+- `POST /ingest`: CSV as raw request body. Rows failing validation (negative values, empty currency, missing/unparseable amount or dates) are still stored, with status `corrupted`. Returns `{ status, ingested, errors, rows }`. No idempotency — re-posting the same file appends duplicates.
+- `GET /invoice/:id`: invoice (with status) + its payments (with PSP-owned status). Resolves surrogate id or invoice_number (newest wins on duplicate numbers).
+- `POST /payment/:invoice_id`: zod-validated `{idempotency_key: uuid, amount, currency}`; only `open`/`partially_paid` invoices are payable (409 otherwise — includes `corrupted`), repeated key replays the original payment (200), else inserts a `pending` payment (201).
+
+Money is stored as integer minor units (cents): CSV `150` → `15000`.
 
 ## Seed data
 
-- customers: `cus_ada`, `cus_grace`
-- plans: `plan_basic` ($10/mo), `plan_pro` ($50/mo)
-- `cus_ada` has an active `sub_1` on pro with open `inv_1`
+Seeded by ingesting `invoices.csv` (25 rows → 22 open + 3 corrupted, 21 customers):
+
+- `INV-1006` (line 7): empty currency code
+- `INV-1010`: negative quantity and amount
+- `INV-1020`: missing amount
+
+Duplicate invoice numbers (`INV-1001` ×2, `INV-1006` reused) are ingested as-is — not a validation rule on the canvas. Customers dedupe by email case-insensitively (`Acme Corp`/`ACME CORPORATION`/`acme corp` → one customer); rows without email get a customer keyed by name.
 
 ## Extension cheatsheet (interview)