HH/reports/feedback-modules-qa-report.md

Feedback Modules QA Audit Report

Modules: Complaint · Inquiry · Observation · Suggestion · Appreciation Date: 2026-06-14 Environment: Local dev (runserver :8000 → PostgreSQL px360_db), headed Chromium (DISPLAY=:0), console email backend, isolated E2E Test Hospital mirroring HH-N (Al Nuzha) hierarchy. Method: Exploratory, run-to-completion Playwright spec (e2e/tests/workflows/feedback-modules-audit.spec.ts); every step guarded; console/page-error/HTTP-error listeners captured observations into e2e/results/audit-observations.json. Roles tested deeply: px_admin, hospital_admin, dept_manager, px_employee, staff; access-reachability for all 10 roles.

Headline numbers (clean run, server stable)

• 26/26 audit tests completed (0 hard failures). 8.1 min.
• 124 observations: 50 PASS · 49 WARN · 10 INFO · 5 SKIP · 0 FAIL (server up)
• Baseline lifecycle specs: inquiry 7/7 pass, observation 5/6 pass, complaint 0/10 (suite broken — see §5).

---

1. Complaint

Routes: public /complaints/public/submit/, list /complaints/, detail /complaints/<id>/, track /complaints/public/track/, export /complaints/export/csv/.

✅ Handled correctly

• Public form submits successfully and creates a Complaint (verified in DB: 4 records in E2E-HOSP).
• Authenticated list renders table + status badges for hospital_admin, dept_manager, px_employee, staff (24 badges each).
• Detail page opens; add-note posts successfully; CSV export returns 200 text/csv (6 lines, valid header).
• Authenticated create form (/complaints/new/) renders.

⚠️ Issues found

• Reference number not surfaced/extractable on the public success page — the test couldn't capture a CMP-… ref from body or URL, so the subsequent track test ran with a dummy ref.
• activate action did not reflect in_progress in the page text after submit (status text present: false).
• px_admin list renders no table (see cross-cutting §4.1).
• One net::ERR_ABORTED on a detail navigation (test re-clicked; benign).

🔧 Improvements

• Display the reference number prominently on the public success page (and in the success URL) so submitters can track.
• Verify the activate/self-assign flow updates the visible status badge immediately.

2. Inquiry

Routes: public /inquiries/public/submit/, list /inquiries/, detail /inquiries/<id>/.

✅ Handled correctly

• Public form submits and creates an Inquiry (3 records in E2E-HOSP).
• List + detail open for hospital_admin, dept_manager, px_employee, staff.
• Respond modal opens for hospital_admin (form present, closeable).
• Status badges render.

⚠️ Issues found

• Respond action only visible to hospital_admin — dept_manager, px_employee, staff see no respond button on inquiry detail. Confirm whether this is intended RBAC or a gap.
• Public submit did not surface an INQ-… reference (regex didn't match the success body) — minor.
• px_admin list renders no table (§4.1).

🔧 Improvements

• Clarify/extend inquiry "respond" permissions to the roles expected to action inquiries.
• Show the inquiry reference on the public success page.

3. Observation

Routes: public /observations/new/, success /observations/submitted/<code>/, track /observations/track/, list /observations/, detail /observations/<id>/.

✅ Handled correctly

• Public form renders, and a valid POST (with incident_datetime in YYYY-MM-DD HH:MM:SS) returns 302 with a tracking code (e.g. OBS-R208ER) — 3 records created in E2E-HOSP.
• Public track page responds.
• List + detail open for hospital_admin, dept_manager, px_employee, staff (all pass).

⚠️ Issues found

• UI cascade is fragile & format-sensitive: the public form's hospital → location_type → area → category → department → section is pure AJAX; submission silently re-renders (200, no redirect) unless incident_datetime includes seconds (2026-06-13 12:00 fails, 2026-06-13 12:00:00 works). A real user using the date picker is fine, but raw entry is unforgiving.
• px_admin list renders no table (§4.1).

🔧 Improvements

• Add a proper datetime picker so the canonical format is always sent.
• Show client-side validation errors clearly when the cascade fields are incomplete.

4. Suggestion (/suggestions/, feedback app)

Routes: public /suggestions/public/suggestion/ (POST-only), list /suggestions/, create /suggestions/create/, detail /suggestions/<id>/.

✅ Handled correctly

• Authenticated list renders for hospital_admin, dept_manager, px_employee, staff.
• Public endpoint correctly returns 405 on GET (POST-only) and 400 on empty POST — i.e. the endpoint is wired and validating.

⚠️ Issues found (highest-impact)

• No public GET form for suggestions — anonymous users have no UI to submit a suggestion (only a POST endpoint). Inconsistent with Complaint/Inquiry/Observation which all have public GET forms.
• Authenticated create did NOT persist a record. The test reported a "success indicator" but 0 Feedback rows exist in E2E-HOSP (only 2 stale rows in HH-N from prior manual testing). The create form (feedback_form.html) fields are comm_req, contact_name, contact_phone, hospital(select), message — no title, category, or feedback_type inputs, yet the test (and the FeedbackForm ModelForm) assume them. The custom feedback_create view reads only those 5 fields. → The authenticated suggestion-create flow needs manual verification; the automated harness could not reliably drive it.

🔧 Improvements

• Add a public GET form for suggestions (parity with other modules), or document that suggestions are internal-only.
• Reconcile the create form's actual fields with the FeedbackForm ModelForm (title/category/feedback_type are model fields but absent from the create UI) and verify an authenticated user can actually save a suggestion end-to-end.

5. Appreciation (/appreciation/)

Routes: public /appreciation/public/submit/ (JSON POST), list /appreciation/, detail /appreciation/detail/<id>/, activate /appreciation/detail/<id>/activate/, send /appreciation/detail/<id>/send/, acknowledge /appreciation/acknowledge/<id>/.

✅ Handled correctly

• Public JSON POST creates a DRAFT (APR-… ref returned, CSRF handled) — 2 records in E2E-HOSP.
• List + detail open for hospital_admin.
• Full internal workflow works: activate (select staff → activated, AI analysis triggered) → send (status → sent, notification dispatched to console). ✅

⚠️ Issues found

• No public GET form — like suggestions, appreciation public submission is POST-only (GET → 405). No anonymous-facing UI.
• acknowledge was not exercised (only the recipient can acknowledge; out of scope for a non-recipient session).
• One net::ERR_ABORTED on a detail navigation (benign test artifact).

🔧 Improvements

• Add a public GET form (parity), or document that appreciation submission is JSON-API-only.
• Add an automated path to test acknowledge (login as the recipient staff's user).

---

6. Cross-cutting findings (prioritized)

🔴 P1 — px_admin sees empty lists across ALL modules

For Complaint, Inquiry, Observation, and Suggestion, px_admin (confirmed logged in, scoped to E2E Test Hospital, is_px_admin()=True) renders no data table, while every other role sees populated tables. Records demonstrably exist in E2E-HOSP. This points to the tenant_hospital session-scoping logic for PX Admins not resolving E2E-HOSP (the list views branch on if user.is_px_admin(): if selected_hospital: …). Needs investigation — PX Admin is the primary oversight role and currently appears to get a blank view.

🔴 P1 — Suggestion authenticated create doesn't persist

See §4. No Feedback record is saved from the authenticated create flow. The create form's fields don't match the model/form assumptions.

🟡 P2 — Inconsistent public-form parity

Module	Public GET form	Public submit
Complaint	✅	✅
Inquiry	✅	✅
Observation	✅	✅
Suggestion	❌ (POST-only)	⚠️ endpoint exists
Appreciation	❌ (POST-only)	⚠️ JSON-only

Anonymous UX is inconsistent. Either add GET forms for the last two or document them as internal/API-only.

🟡 P2 — Inconsistent reference-number formats

CMP-YYYYMMDD-NNNNNN, INQ-<8 hex>, OBS-<6 alnum>, APR-<8 hex>. Consider a unified scheme for tracking/UX consistency.

🟢 P3 — Reference numbers not surfaced on success pages

Complaint & Inquiry public success pages don't expose the reference in an easily extractable way (affects tracking UX and the public track flow).

🟢 P3 — Inquiry "respond" limited to hospital_admin

Other actioning roles (dept_manager, px_employee, staff) see no respond button. Confirm intent.

🟢 P3 — Observation public form date format & cascade fragility

Requires incident_datetime with seconds; cascade gives no inline validation feedback.

⚪ Ops — Dev server stability under automation

During one run the runserver autoreloader restarted mid-suite (connection reset → refused), invalidating that pass. For repeatable automation, run a dedicated QA server with --noreload (or a gunicorn/uvicorn instance) instead of the autoreloading dev server.

---

7. Role / access matrix (observed)

All 10 roles were able to reach every module root (no incorrect login-redirect blocks). The audit's PASS/WARN labels for the matrix are spurious (a comparison-flaw in the exploratory status logic: onLogin === expected mixed "blocked?" with "expected-allowed?"), so the matrix should be treated as inconclusive for fine-grained RBAC. The documented e2e/tests/roles/access-matrix.spec.ts remains the source of truth for reachability; this audit did not contradict it. Recommended: fix the audit's matrix status comparison and re-run for a clean RBAC signal.

---

8. Baseline lifecycle comparison (existing specs)

Spec	Result	Notes
complaint-lifecycle.spec.ts	0/10 run	BROKEN — uses outdated #id_complainant_name / #id_… selectors; the public complaint form now exposes fields by name= only. Test 1 fails → 9 skipped (serial). Update selectors.
inquiry-lifecycle.spec.ts	7/7 pass	✅ Healthy.
observation-lifecycle.spec.ts	5/6 pass	1 fail: "severity badges render correctly". Public submit works here.

The complaint-lifecycle staleness is itself a finding — the public complaint form's field IDs were removed but the test wasn't updated.

---

9. What was changed for this audit (all reversible)

• New: apps/core/management/commands/create_e2e_isolated_env.py — builds isolated E2E-HOSP (mirrors HH-N: 27 Areas, 37 Departments, 171 Sections, 1 SubSection) + 10 role users.
• New: e2e/tests/workflows/feedback-modules-audit.spec.ts + e2e/helpers/audit.ts.
• Edited: .env → EMAIL_BACKEND=console (backup at .env.bak_qa); playwright.config.ts → env-overridable timeout/actionTimeout/navigationTimeout/slowMo (defaults preserve prior behavior).
• Created: isolated hospital E2E-HOSP (kept for future QA), org E2E-ORG, source E2E-TEST, 10 e2e-*@px360.test users (password Dev@123456), plus audit-generated records inside E2E-HOSP.
• No migrations, no commits.

10. Reproduce

# 1. (one-time) build the isolated sandbox mirroring HH-N
uv run manage.py create_e2e_isolated_env --delete-existing
# 2. email -> console (already applied; .env.bak_qa has the SMTP original)
# 3. run the audit headed
E2E_TIMEOUT=120000 E2E_ACTION_TIMEOUT=12000 E2E_NAV_TIMEOUT=20000 E2E_SLOWMO=60 \
  npx playwright test e2e/tests/workflows/feedback-modules-audit.spec.ts --workers=1 --headed
# 4. observations + report inputs
cat e2e/results/audit-observations.json

11. Recommended next steps (for these 5 modules)

1. Fix P1 px_admin empty-list scoping.
2. Fix P1 suggestion authenticated create persistence; reconcile form fields.
3. Decide public-form parity for suggestion & appreciation (add GET forms or document as internal).
4. Update complaint-lifecycle.spec.ts selectors (name= instead of #id_).
5. Surface reference numbers on all public success pages.
6. Re-run this audit after fixes (sandbox is retained); then proceed to the next module group.

---

12. Champion / Manager Workflow Audit (send-to-department lifecycle)

Covers the full multi-actor loop the original audit missed: PX-team sends → champion investigates → manager reviews → PX-team accepts → resolve, plus the token investigation sub-flow and both rejection loops. Spec: e2e/tests/workflows/champion-manager-workflow.spec.ts. State asserted after every step via the get_e2e_workflow_state CLI helper (Node can't read the DB).

Result: 4/4 flows pass. 32 observations: 26 PASS · 1 FAIL · 3 WARN · 2 INFO. Two real bugs were found and fixed during the run.

✅ Handled correctly (verified end-to-end)

Flow	Path	Result
A — happy path	PX send → champion response → manager approve → PX accept → resolve	✅ all 11 state transitions correct (sent → response_submitted+pending → approved → acceptable → resolved)
B — token investigation	champion creates Qs (token) → staff answers (token) → champion reviews + writes response	✅ response_submitted=True, manager_review_status=pending
C1 — manager reject loop	manager rejects → response cleared + response_submitted=False → champion re-responds → approve → accept → resolve	✅ reject correctly returns the complaint to the champion
C2 — PX not-acceptable loop	PX marks not-acceptable → response cleared → champion re-responds → approve → accept → resolve	✅ correctly returns to champion

The ComplaintInvolvedDepartment state machine behaves exactly as designed: sent → response_submitted(pending) → approved → acceptable, and both rejection branches reset response_submitted=False + clear the response, sending it back to the champion.

🔧 Bugs found & fixed during this audit

1. NameError: name 'InvestigationAnswer' is not defined at apps/complaints/views.py:4021 — champion_start_investigation used InvestigationAnswer without importing it, so the champion's "create questions" POST returned 500 (DB writes before the crash still persisted, masking the error). Fixed: added InvestigationAnswer to the view's imports.
2. AttributeError waiting to happen at apps/complaints/views.py:4064 — staff_member.phone_number, but the Staff model field is phone. Would have crashed the moment the NameError above was fixed. Fixed: phone_number → phone.

⚠️ Remaining issues

• ReferenceError: lucide is not defined (frontend, Flow B token pages) — the icon library (lucide) isn't loaded on the investigation templates (investigation_questions.html / investigation_respond.html / investigation_review.html). Cosmetic only; the forms still work. Fix: include the lucide script on those templates (or use the icon partial the rest of the app uses).
• GET /organizations/departments/<id>/analytics/ → ERR_ABORTED (Flows C1/C2) — the department-analytics XHR on the dept pages fails (likely 404/500). Minor; doesn't block the workflow.
• SMS gateway (Mshastra) "IP address not allowed" — external provider config, not app code; expected in dev.

🔩 Test harness added

• apps/core/management/commands/create_e2e_isolated_env.py extended — binds e2e-dept-manager as department.manager and creates a Staff profile for e2e-staff in the champion's department (Contact Center), so all three workflow roles exist.
• apps/core/management/commands/seed_e2e_complaint.py (new) — seeds an open complaint + accused staff, prints ids.
• apps/core/management/commands/get_e2e_workflow_state.py (new) — prints the full workflow state (idept fields + tokens) for assertions.
• e2e/tests/workflows/champion-manager-workflow.spec.ts (new) — Flows A/B/C1/C2, drives the real HTTP endpoints (AJAX + form POSTs), asserts state after each step.

Reproduce

uv run manage.py create_e2e_isolated_env --delete-existing
E2E_TIMEOUT=120000 E2E_ACTION_TIMEOUT=12000 E2E_NAV_TIMEOUT=20000 \
  npx playwright test e2e/tests/workflows/champion-manager-workflow.spec.ts --workers=1 --headed

---

13. Inquiry & Observation dept-response workflow (simpler flow)

These two modules use a simpler flow than complaints: PX-team sends to department → champion responds → PX reviews (accept/reject). No manager review, no investigation questions. State is flat fields on the Inquiry/Observation model (no join table). Spec: e2e/tests/workflows/inquiry-observation-dept-workflow.spec.ts, 6 flows (3 per module), driven via the real HTTP endpoints with state asserted via get_e2e_dept_response_state.

Result: 6/6 flows pass. 59 observations: 50 PASS · 5 WARN · 0 FAIL.

✅ Handled correctly (verified)

Flow	Inquiry	Observation
A — happy path (send → champion response → PX accept → resolve)	✅	✅ (resolve best-effort — see below)
B — token path (send mints token → champion token response → PX accept)	✅ backend	✅ backend
C — reject loop (PX not-acceptable → response cleared → champion re-responds → accept → resolve)	✅	✅

State machines behave correctly: send sets transferred_to_department (inquiry) / sent_to_department=True (observation); champion response sets department_response_en + department_responded_at + dept_response_acceptance_status=pending; PX acceptable → acceptance_status=acceptable; PX not_acceptable → clears the response and returns it to the champion (verified — response_en_set=False after reject). Inquiry resolve requires a PX response first (inquiry_respond) — handled.

⚠️ Issues found → ✅ fixed (re-verified, 0 FAIL)

1. Token-response form page is broken (both modules) — ✅ FIXED. Root cause: the 8 public/token templates {% extends "template.html" %}, a static dashboard with no {% block content %}, so the form block was dropped. Fix: changed all 8 to {% extends "layouts/public_base.html" %}. Re-verified: B-token-form-render now PASS (form + csrf render), anonymous token POST works.
2. NameError: get_email_header_html during inquiry transfer — ✅ FIXED. Root cause: apps/notifications/settings_service.py imported NotificationService but not get_email_header_html. Fix: added get_email_header_html to the lazy imports. Re-verified: no NameError in the re-run.
3. Inquiry dept-response page 500 (DoesNotExist) — ✅ FIXED. Root cause: templates/complaints/inquiry_department_response.html was missing. Fix: created it (mirrors the observation template). Re-verified: GET renders, no 500.
4. lucide is not defined (icons broken) — ✅ FIXED. Root cause: unguarded lucide.createIcons() in layouts/base.html:157 threw when lucide hadn't initialized during rapid navigation. Fix: guarded that call + added a guarded init to layouts/public_base.html.
5. /organizations/departments/<id>/analytics/ XHR failure — ✅ not a bug. Endpoint returns 200; the ERR_ABORTED was a test-navigation artifact.

⚠️ Remaining (behavior, not bugs)

• Observation can't jump new → resolved — its status machine requires intermediate steps (new → triaged/assigned/in_progress → resolved). observation_change_status accepts the POST (302) but the status stays new. Confirm whether the dept-response flow should drive the status to a resolvable state (e.g., auto-activate on send). Reported as WARN, not FAIL.

🔩 Test harness added

• apps/core/management/commands/seed_e2e_dept_response.py — seeds an open inquiry/observation, prints ids.
• apps/core/management/commands/get_e2e_dept_response_state.py — prints the dept-response state + token for assertions.
• playwright.config.ts — E2E_MAXIMIZED=1 launches the browser maximized/fullscreen (viewport=null + --start-maximized).

Reproduce

E2E_MAXIMIZED=1 E2E_TIMEOUT=120000 E2E_ACTION_TIMEOUT=12000 E2E_NAV_TIMEOUT=20000 E2E_SLOWMO=70 \
  npx playwright test e2e/tests/workflows/inquiry-observation-dept-workflow.spec.ts --workers=1 --headed