The portal had no way to obtain a verified principal: /api/0pt/* requires a
Bearer token, but the browser only had the dev-only HMAC cookie.
- GET /api/auth/login: Authorization Code + PKCE against Zitadel.
- GET /api/auth/callback: code exchange, id_token verification (issuer,
audience, nonce) and a signed session cookie with sub + tenant.
- middleware accepts the signed session as a verified principal; Bearer wins.
- dev login POST is refused in production.
- Sign-in link in the portal shell.
Tenant isolation (security-critical): every app/api/0pt/** route reading
tenant_id from client-controlled query/body never cross-checked it against
the token middleware.ts already verified (x-0pt-verified-tenant). Adds a
shared assertTenantMatchesVerified guard and wires it into all 20+ affected
routes. Also closes two full-leak-by-omission bugs found in this pass
(documents/capture and lakehouse/mappings GETs returned every tenant's rows
when tenant_id was simply omitted) and two "no tenant check at all" reads
(rate-policies/[id], lakehouse/mappings/[id]) reachable by ID alone.
Duplicate-delivery idempotency (correctness-critical): POST /api/orders now
honors Idempotency-Key (sql/006, replay-or-409-on-conflict, Postgres +
in-memory); the Python orchestrator's POST /orders/{id}/run — the actual
delivery trigger — is now idempotent under an asyncio lock, closing a real
concurrent-request race, not just the sequential-retry case.
Engine-duplication resolution: adaptive_router.py's docstring wrongly
invited a bypass ("see b0ts for scraping") contradicting the product's own
compliance stance; corrected, and given real (additive, not escalating)
block-signal tracking so a rate-limited/blocked source is excluded from the
plan instead of hammered. Also fixes a previously-undiscovered production
bug: the Python engine's only call into the TS quality bridge
(X-Internal-Token) was unconditionally 401'd by middleware.ts, which never
checked that credential — the advisory quality bridge was unreachable in any
environment enforcing auth.
Restore drill: executed for real against throwaway Docker Postgres
containers (not simulated). Found and fixed two real blockers along the way:
sql/004 referenced opt_datasets, which no migration ever created (chain was
never actually runnable end to end); lib/db.ts hardcoded ssl:"require",
making any non-TLS Postgres (incl. compose.dev.yaml's own local-dev
container) unreachable. After both fixes: full migration chain applies
cleanly, pg_dump/pg_restore round-trip is byte-identical (row counts +
content checksums), and real application code (lib/orders-store.ts,
SourceRegistryService's Postgres repository) reads the restored data back
correctly, including cross-tenant isolation still holding.
82 new tests (55 TS + 27 Python), all passing; full existing suite (291 TS +
55 Python) still green; typecheck clean on every changed/added file.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
BACKLOG P0-3: /api/health only ever checked *presence* of DATABASE_URL/
REDIS_URL and always returned ok:true regardless of whether anything was
reachable -- ARCHITECTURE.md's own words, "an uptime monitor watching HTTP
200 would show green for a dying instance".
lib/health-checks.ts adds real dependency probes:
- db: a live `select 1` round trip (not just DATABASE_URL presence)
- engine: a real HTTP call to the FastAPI engine's own GET /health
- stripe: Stripe's documented lightweight balance.retrieve() liveness call
- queue: honestly reports "not_integrated" when REDIS_URL is set but
unused -- no Redis client is wired anywhere in this repo, so reporting
"ok" would fabricate a check that doesn't exist
/api/health stays a liveness check (always 200 while the process can
respond; the body's `ok` field is the honest signal). New GET /api/ready
fails closed with 503 when the load-bearing dependency (db) isn't both
configured and reachable -- the "no /ready route" gap.
contracts/openapi.yaml, docs/architecture/ARCHITECTURE.md, and docs/slo.md
updated to match. 17 new unit tests (dependency-injected fakes, no real
network/DB) + 5 route-level tests.
The auth seam read `process.env.AUTH_ENFORCE === 'true'`, so an absent env
var meant NO authentication: any fresh container, local run, DR restore or
redeploy that dropped the variable served /api/0pt/* anonymously, taking the
tenant from the caller-supplied X-Tenant-Id header. 0pt.app is deployed and
was answering anonymously (verified: an invalid Bearer token plus an
arbitrary X-Tenant-Id returned 200), so this was a live hole, not latent.
* authEnforced() now returns true unless AUTH_ENFORCE is explicitly "false"
(trimmed, case-insensitive). Unset/empty/misspelled => ENFORCED.
* assertAuthConfigSafe() throws AuthMisconfigurationError when
NODE_ENV=production and enforcement resolved off, naming AUTH_ENFORCE. It
runs at middleware module load (startup) and on the env-driven request
path, so a production instance can neither boot nor serve a single
unauthenticated request with auth switched off.
* .env.example no longer teaches AUTH_ENFORCE=false; README's Deploy section
states the fail-closed posture (no DEPLOY.md in this repo).
* 7 new tests, incl. the regression case: with AUTH_ENFORCE unset and no
`enforce` option, an unauthenticated request is rejected with 401. No test
was inverted or deleted -- none asserted the open default; the three
enforce:false cases now describe the explicit local-dev escape hatch.
Tests: 121 passed / 1 skipped before, 128 passed / 1 skipped after.
Typecheck and build clean; advisory lint unchanged (16 pre-existing problems).
Scope: auth default only. The matcher still covers /api/0pt/* alone, and the
handlers still take tenant_id from the query/body rather than the verified
principal -- both described in the PR body, neither fixed here.
admin
