Templates reference public/img/logo.png (favicon, brand logo, social
share images) which isn't present in the repo. With the strict manifest
storage this raised ValueError at request time -> HTTP 500. Override
stored_name (and set manifest_strict=False) to fall back to the plain
name so a missing static degrades to a broken asset instead of a 500,
matching the previous non-manifest behaviour.
Co-authored-by: Cursor <cursoragent@cursor.com>
The production .env stores DJANGO_ALLOWED_HOSTS as a JSON array (legacy
format), but env_list only split on commas, yielding broken entries like
'["aimloperations.com"' and causing DisallowedHost (HTTP 400) for every
request. Parse JSON arrays as well as comma-separated values.
Co-authored-by: Cursor <cursoragent@cursor.com>
WhiteNoise's manifest storage strictly resolves every referenced file
during collectstatic, including sourceMappingURL comments in vendored
JS bundles. A missing .map (financial/js/.../dashboard-free.js.map)
broke the prod container at startup. Add TolerantManifestStaticFilesStorage
which leaves unresolved references untouched instead of raising.
Co-authored-by: Cursor <cursoragent@cursor.com>
Replace hardcoded settings with environment-driven config, add Docker
compose for local and production deploys, migrate from pip to uv, and
split Gitea workflows so PRs run tests only while master pushes deploy.
## Summary
- Adds `UserProfile` model with mutually exclusive **Employee** / **Client** types
- Replaces auto-Employee signal with auto-Client profile on user creation
- Data migration: users with time log entries → Employee; others → Client (orphan Employee rows removed)
- Admin UI at `/financial/manage_users` to set any user's type; profile page shows current type
- **Employees** can log time; **Clients** get read-only access to reports and time logs
- Time logs, reports, and dashboard filter to employees only
- 14 new tests covering signals, type switching, access control, and filtering
## Design decisions (from issue Q&A)
1. Client login = read-only financial access (reports + time logs, no edit/log time)
2. Employee and Client are strictly mutually exclusive
3. Admins (superusers) can change type via Manage Users
4. Bulk migration applied for existing users
## Test plan
- [x] `python manage.py test financial.tests` (14 tests pass)
- [x] `python manage.py test public.tests` (21 tests pass)
- [ ] Run migration on staging: `python manage.py migrate`
- [ ] Verify admin can set user types at `/financial/manage_users`
- [ ] Verify employee can log time at `/financial/timekeeping`
- [ ] Verify client sees reports/time logs read-only, cannot log time
- [ ] Verify employee filter dropdown excludes clients
Closes#14
Reviewed-on: #15
## Summary
Follow-up to PR #12 / Lighthouse agentic-browsing audit on `/contact` (score 0.88).
- Add `toolparamdescription` to all contact form fields for WebMCP schema validity
- Move reCAPTCHA outside the annotated `<form>` so `g-recaptcha-response` is excluded from the declarative tool schema
- Always render `toolname` / `tooldescription` on the contact form (not gated by `WEBMCP_ENABLED`)
- Override `django_recaptcha/widget_v3.html` to associate captcha with `form="contact-form"`
- Update docs with HTTPS Lighthouse commands
## Test plan
- [x] `python manage.py test public.tests` — 21 tests pass
- [ ] Deploy and re-run Lighthouse agentic-browsing on `https://aimloperations.com/contact`
- [ ] Confirm `webmcp-schema-validity` passes (expected overall score 1.0)
Reviewed-on: #13
## Summary
- Add `WEBMCP_ENABLED` setting and guarded `webmcp-tools.js` loader on public pages
- Register navigation tools (`list_services`, `get_page_content`, `navigate_to_service`, `open_contact_with_subject`) and contact tool (`submit_contact_inquiry`)
- Annotate contact form with declarative WebMCP attributes for Lighthouse form coverage
- Extend `PUBLIC_PAGE_ENTRIES` with summaries shared by SEO and WebMCP tools
- Add `docs/webmcp.md`, update `docs/agentic-browsing.md`, and regression tests
Closes#9
## Test plan
- [x] `python manage.py test public.tests` — 20 tests pass
- [ ] Set `WEBMCP_ENABLED=True`, enable Chrome experimental web platform features flag
- [ ] Verify homepage HTML includes `webmcp-config` and `webmcp-tools.js`
- [ ] Verify `/contact` form has `toolname="submit_contact_inquiry"`
- [ ] Run Lighthouse agentic-browsing on `/` and `/contact` with experimental flag
- [ ] Call `list_services` from WebMCP-capable Chrome on homepage
- [ ] Call `submit_contact_inquiry` on `/contact` with `DEBUG=True` (reCAPTCHA bypass)
Reviewed-on: #12
Load analytics on page load so visitors who never interact still produce data, while an informational banner lets users acknowledge the practice. Migrate legacy declined and accepted consent values so prior opt-outs are cleared and tracking resumes.
Closes#7
Reviewed-on: #10
## Summary
- Add `robots.txt`, `sitemap.xml`, and `llms.txt` endpoints so crawlers and AI agents can discover public pages
- Fix accessibility tree issues: contact form labels, semantic nav/profile dropdown buttons, cookie consent dialog ARIA, and dead `href="#"` links
- Reduce layout shift on homepage/contact via hero min-heights, trusted-by marquee sizing, and fixed cookie banner
- Add regression tests and `docs/agentic-browsing.md` for Lighthouse agentic-browsing audits
Closes#5. Also addresses overlap with #3 (sitemap + robots.txt).
## Test plan
- [x] `python manage.py test public.tests` (14 tests pass)
- [ ] Verify `GET /robots.txt`, `/sitemap.xml`, `/llms.txt` return 200 in staging/prod
- [ ] Confirm contact form labels visible and submittable on `/contact`
- [ ] Run Lighthouse `--only-categories=agentic-browsing` on homepage and contact page
- [ ] Smoke test mobile nav dropdown and cookie consent banner
Reviewed-on: #8