Team decision: production DB connection and first-boot owner email are hardcoded in the app instead of a .env file, matching the box's other apps. Gated to NODE_ENV=production so dev/tests keep using SQLite (389 tests green). Committed value is the CHANGE_ME_PASSWORD placeholder; the real password is set only on the box, never pushed. server.ts: HARDCODED_DATABASE_URL. seed.ts: OWNER_EMAIL default. docker-compose.yml: drop env_file. docs: PATHS-AND-NAMES.md + .env.example updated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Red-team highs/mediums (auth surface):
- staff_user gains failed_count/locked_until; login() locks an account for 15min
after 8 consecutive failures, clears the counter on success (both engines).
- login() is now constant-time: a missing email still spends one scrypt verify
against a fixed dummy hash — no timing/enumeration oracle. Returns a typed
{ok:false,reason} instead of null.
- POST /auth/login gets per-IP (20/min) AND per-email (8/15min) fixed-window
rate limiting layered over the lockout; 429 on trip.
- Credential reveal routes (db-password, module password, module secret) get a
per-user reveal limiter (20/min) so one insider cannot bulk-decrypt everything;
added missing getClientModule 404 guards.
- app.set(trust proxy,1) so req.ip is the real client behind nginx/Caddy and the
per-IP limiters actually bucket per client.
- makeRateLimiter extracted to rate-limit.ts (shared, no server<->api cycle).
Tests: lockout, counter-reset, unknown-email, and existing auth all green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
startServer picks Postgres when DATABASE_URL is set, SQLite file otherwise.
New pg-engine.test.ts (opt-in via HQ_PG_TEST_URL, wipes its target schema):
migrations from zero, invoice issue with series numbering + due-date stamp,
paisa-exact payment to paid, exactly-once reminder upsert via ON CONFLICT,
savepoint-nested rollback (incl. the mixed dbPassword+patch write), and a
full HTTP boot smoke (health/login/clients) on the postgres engine.
Verified locally against PostgreSQL 17: 5/5. Default suite (sqlite)
346/346, typecheck clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
P1: every repo, route, scheduler path and test now speaks one async DB
interface (all/get/run/exec/transaction). better-sqlite3 stays underneath
for dev/tests via SqliteDb (statement cache, manual BEGIN/SAVEPOINT
nesting); openDb stays sync so fixtures remain one-liners. Zero behavior
change: 346/346 tests, typecheck clean.
P2: migrations-pg.ts — numbered Postgres migration set (001-init: bigint
paise, integer flags, TEXT ISO dates, TEXT JSON; reminder CHECK born in
final widened form) + schema_migrations runner.
P3: db-pg.ts — pg Pool behind the same interface: ?->$n rewrite outside
string literals, transactions pinned to one pooled client via
AsyncLocalStorage with savepoint nesting, int8/numeric parsed to JS
numbers so paise math is identical on both engines.
Also: reminder upsert INSERT OR IGNORE -> ON CONFLICT DO NOTHING
(portable, same changes-count semantics).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Mount the one unauthenticated read route OUTSIDE the /api auth router.
A live token resolves via validateShare to exactly one document, rendered
inline (application/pdf) through the existing documentHtml -> renderPdf path;
unknown/expired/revoked tokens return a plain 404 "link expired or invalid"
HTML page carrying no document data. Adds a dependency-free per-IP fixed-window
rate limiter to blunt token brute-forcing. renderPdf is injected so the route
is TDD-tested without launching Chrome. Public reads write no audit; the token
is never logged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Thomas Joise