CT115 (Mumuni relay #747) — two live-instance defects after PR #12:
1. GET /api/whatsapp/mock-log 500'd with a valid admin token:
'no such column: whatsapp_log.message_text'. The model gained
message_text/wa_message_id/ticket_number (and dropped command) in
4afdc36 with no migration, so legacy DBs keep the (command, ...) shape.
ensure_legacy_schema (startup, app/main.py) now adds the three missing
columns idempotently and backfills legacy command bodies into
message_text before dropping the obsolete NOT NULL command column, so
both the mock-log read path and the ORM write path work on healed DBs.
New producer/consumer regression (tests/test_whatsapp_log_legacy_heal.py)
reproduces the exact OperationalError, then asserts 200 + data.
2. ROLE_ALIASES gap: underscore legacy roles (cs_rep, cs_manager,
fm_dispatcher) were not mapped, so normalize_legacy_user_roles could not
converge rows like user 18 (test@denya.com, role 'cs_rep') and the
frontend stranded them on /tickets. Added the underscore aliases; tests
assert normalize_role('cs_rep') == 'CS Rep' and a cs_rep row converges
and authenticates.
The P0 templates loaded Alpine.js from cdn.jsdelivr.net and Tailwind from
cdn.tailwindcss.com, so LAN-only demo clients got a login page whose JS never
engaged (stuck form). Vendor both libraries under app/static/vendor/
(alpine-3.17.2.min.js, tailwind-3.4.17.js) served same-origin at /static,
point base.html at local paths, and drop both CDN hosts from the CSP
(script-src/style-src stay 'self' 'unsafe-inline'; connect-src 'self').
HTML pages now ship Cache-Control: no-cache; vendored assets are cached
public, max-age=31536000, immutable (versioned filenames). HSTS stays
TLS-gated. Adds tests/test_frontend_vendoring.py (no external script src on
/login, both vendor paths 200, no-cache + immutable header checks, CSP
without CDN hosts).
- Remove POST /api/auth/register (404); no sign-up UI; users are admin-managed
- Add admin-only POST/PATCH/DELETE /api/auth/users (forced canonical roles,
self-lockout + reference guards)
- Unify role model in app/core/roles.py; reject unknown roles at creation and
at login/JWT validation; startup normalizes unambiguous legacy aliases
- Login rate limiting ~5 fails/15 min per IP+email -> 429 (in-process, tunable)
- WhatsApp webhook requires X-Webhook-Secret; fail-closed when env unset;
GET handshake uses constant-time verify token (403 on mismatch)
- GET /api/whatsapp/mock-log now requires auth
- Security headers middleware: X-Frame-Options DENY, nosniff, CSP on HTML,
HSTS behind TLS
- Pagination: limit alias for page_size, hard cap enforced, both -> 422
P0.1 — fail-closed secrets:
- config.py: no default SECRET_KEY; refuses to boot when unset, a known
placeholder, or <32 chars. Generate with: openssl rand -hex 32.
- docker-compose.yml: literal secrets removed; runtime env now comes from
a git-ignored .env via env_file. .env.example added as template.
- .gitignore already covers .env (verified).
P0.2 — locked CORS:
- main.py: CORS_ORIGINS must be an explicit comma-separated allow-list.
'*' or an empty value refuses to boot (was: silently ['*'] with
allow_credentials=True).
P0.3 — role-safe registration:
- services/auth.py: client-supplied 'role' is IGNORED on POST
/api/auth/register; self-registered users always get the
least-privilege 'CS Rep' role. Unauthenticated callers can no longer
mint Admin/Jerome, Admin/Wahab, or Director accounts.
Tests:
- conftest.py sets test SECRET_KEY/CORS_ORIGINS before app import.
- New tests/test_p0_hardening.py (8 tests): role-escalation blocked for
Admin/Jerome and Admin/Wahab, duplicate-email 409, and subprocess
boot-validation for placeholder/short/missing secret + wildcard CORS.
- Full suite: 44 passed.
Redeploy note (per research): seed_units/seed_categories are insert-only,
so the Aug-26 redeploy does NOT orphan historical tickets referencing
units 103E/103W/105E/105W or the legacy 34-category tree. Pending
Wahab: are 103E/103W/105E/105W real apartments dropped from the Excel
regeneration? Optional follow-up: floor-number backfill for already-
seeded units (mapping corrected floors; existing rows keep old values).
Checks per HARDENING.md acceptance:
- [x] starting without a real key fails loudly (subprocess-verified)
- [x] compose carries no literal secret; secrets come from .env
- [x] CORS_ORIGINS explicit allow-list, '*' rejected
- [x] unauthenticated register cannot mint Admin/* or Director
- Add Cancelled as terminal status reachable from all active states; exclude
from SLA breach reporting and dashboard active counts; add to status pickers
- Persist customer phone on ticket create/update (was silently dropped);
add tickets.phone migration + legacy self-heal guard
- Add admin-only DELETE /api/tickets/{id} (removes timeline/photos/escalations)
- Add GET /api/auth/users for the assign-technician dropdown (was hardcoded)
- TicketOut now returns nested unit/category so the detail page stops showing
'—' for Unit/Property/Category
- Ticket numbering uses max+1 so deletions never re-issue a number
- New-issue form: require Category and (standard mode) Priority client-side
- Tests: 11 new cases covering cancellation, SLA exemption, phone, delete,
users endpoint, numbering
- Ticket CRUD with full 15-status lifecycle enforcement
- Auto-generated ticket numbers (PAV-YYYY-NNNNN)
- Status transitions validated and auto-logged to timeline
- Reopen with 7-day window + auto-escalation to Ama
- SLA engine: priority-based deadlines and breach detection
- Photo uploads to uploads/ with multipart support
- Category system seeded from PRD §7 taxonomy (20 top-level)
- Flat and tree category listing endpoints
- SLA status check endpoint (/api/tickets/{id}/sla)
- Paginated ticket listing with filters
root