No description
Find a file
DJP 6320fb389c style: dark slate theme matching the original SPA's look
Pure styling sweep — no behaviour changes, no new dependencies, all
data-tutorial-id selectors preserved. typecheck / lint / build clean.
Main entry 16.36 KB gz (was 12.85), recharts unchanged at 154.57 KB gz.

Foundations:
- body bg-slate-950 text-slate-100, color-scheme:dark, Apple-system font.
- .btn-primary indigo, .btn-secondary slate-700, .card slate-900 +
  slate-800 border, .input slate-800/600, .label slate-400 uppercase.

Top chrome:
- Navbar rewritten as a two-row header: title row with brand + dynamic
  subtitle (filename + row count once a timelog is loaded), then the
  three coloured upload pills inline, then user identity + Sign-out text
  link. Tab row hangs off `border-b border-slate-800` with the
  original's rounded-top "tab" look — active `bg-slate-800 text-white
  border-t border-x border-slate-700`, inactive `text-slate-400`.
- HeaderUploads now matches the original: indigo Time Log pill, emerald
  Deliverable when loaded, violet Project Summary when loaded; neutral
  slate-700 surfaces when empty. Re-uploading replaces (no Clear
  button). Errors collapse to a small ⚠ marker so the header height
  doesn't jump.
- StatsBar: bg-slate-900/50 strip below the navbar with five stacked
  stats (label slate-400 / value white).

Filter rail:
- bg-slate-900/30 strip with stacked-label fields and slate-800/600
  inputs. Native <select multiple> capped to h-20 for the brand/division/
  hub/userRole multiselects (closer to the original's selects than our
  prior chip pattern). Reset is a slate-400 text link, pushed right.
  Filter blocks only render when their option list is populated.

Charts (all six):
- CartesianGrid stroke #334155, axis ticks #94a3b8 on #475569 axis lines.
- Tooltip contentStyle: slate-800 surface + slate-700 border + slate-200
  text, indigo cursor highlight. Tooltip filterNull preserved on Project
  Load.
- "Available"/"Idle" greys swapped to slate-600 so they're visible on
  dark. Primary booking blue swapped to indigo so it harmonises with the
  rest of the indigo accent set.

Side panel + FAB:
- ChatView panel slate-900 / slate-700 border, indigo user bubbles,
  slate-800 assistant bubbles. Repositioned to bottom-24 right-6 so it
  clears the FAB. ChatToggle now indigo, moved to bottom-6 right-6.

Pages:
- Login: slate-950 page, slate-900 card, indigo affordances.
- Department / Resourcing / Bookings / Forecast / ProjectTypeSummary /
  TimeLogDetail / Tutorial all recoloured. Forecast's capacity-decision
  banner is inlined rather than using .card so it can carry the
  colour-coded tint. Bookings virtualised table re-skinned to slate-800
  header / slate-300 rows.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-17 22:28:41 -04:00
backend fix: stop flooding Upload banner with every Zoho column we don't use 2026-05-17 21:52:04 -04:00
deploy deploy.sh: fix self-collision in slug check when clone path != slug 2026-05-16 13:47:03 -04:00
frontend style: dark slate theme matching the original SPA's look 2026-05-17 22:28:41 -04:00
.env.example feat: Forecast, Project Type Summary, Time Log Detail, AI Chat, filters v2, stats bar, RBAC 2026-05-17 21:40:03 -04:00
.gitignore Initial commit: dockerised FastAPI backend + React/Vite frontend rewrite 2026-05-16 12:37:04 -04:00
docker-compose.yml Initial commit: dockerised FastAPI backend + React/Vite frontend rewrite 2026-05-16 12:37:04 -04:00
README.md Initial commit: dockerised FastAPI backend + React/Vite frontend rewrite 2026-05-16 12:37:04 -04:00

utilisation-dept

L'Oréal Utilisation Dashboard — internal tool that merges Zoho time-log exports with Airtable resource/booking data and renders utilisation charts per department, per person, and per project.

This is a clean rewrite of an earlier static SPA that shipped an Airtable Personal Access Token in the JS bundle. The new architecture keeps all secrets on the backend.

Architecture

Browser ─► Apache (optical-dev.oliver.solutions, shared vhost)
            ├─ /utilisation-dept/         → static SPA (Vite build in /var/www/html/utilisation-dept/)
            └─ /utilisation-dept/api/     → FastAPI container (127.0.0.1:<port>)
                                                │
                                                └─ Airtable REST API (PAT held in .env)
  • Backend: FastAPI + uvicorn in a Docker container, bound to 127.0.0.1 only. Apache fronts the public traffic.
  • Frontend: React + Vite + TypeScript + Tailwind + Recharts. Vite base: '/utilisation-dept/'. Built into /var/www/html/utilisation-dept/.
  • Database: none. Airtable is the source of truth; uploaded Zoho files are parsed in memory and discarded.
  • Auth: local admin account today (bcrypt creds in .env). Designed for an Azure AD/MSAL swap-in later — see app/auth/azure.py.

First-time setup

1. Rotate the Airtable PAT (do this before anything else)

The old SPA had its PAT hardcoded in the JS bundle. Assume that token is compromised.

  1. Go to https://airtable.com/create/tokens.
  2. Revoke the old token (patHAB...) if it still exists.
  3. Create a new Personal Access Token:
    • Name: utilisation-dept-backend
    • Scopes: data.records:read
    • Access: limit to base appoByydxIQANKtSh only, tables Resource and Booking Resource.
  4. Copy the token. You'll paste it into .env below.

2. Generate secrets

# Session secret (paste output into SESSION_SECRET):
python3 -c "import secrets; print(secrets.token_urlsafe(48))"

# Admin password hash (replace 'your-password' with the real password):
python3 -c "from passlib.hash import bcrypt; print(bcrypt.hash('your-password'))"

Heads-up: bcrypt hashes contain $ characters. Wrap the hash in single quotes in .env, otherwise docker-compose will treat the $... segments as variable substitutions and the container will receive a truncated value:

ADMIN_PASSWORD_BCRYPT='$2b$12$abc...xyz'

3. Configure .env

cp .env.example .env
# edit .env and fill in:
#   AIRTABLE_PAT          (from step 1)
#   SESSION_SECRET        (from step 2)
#   ADMIN_PASSWORD_BCRYPT (from step 2)

Local development

# Backend
cd backend
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
uvicorn app.main:app --reload --port 8200

# Frontend (in another terminal)
cd frontend
npm install
npm run dev
# open http://localhost:5173/utilisation-dept/

The Vite dev server proxies /utilisation-dept/api/* to http://localhost:8200.

To skip auth entirely during dev, set DEV_AUTH_BYPASS=true in .env (NEVER do this in production — the startup banner will warn loudly).

Local Docker

docker compose up --build
# health: curl http://localhost:8200/api/health

Tests

cd backend && pytest
cd frontend && npm run typecheck && npm run lint

Deployment to optical-dev.oliver.solutions

The server hosts many small apps behind a shared Apache vhost. This app lives at /opt/utilisation-dept/.

First deploy

ssh user@optical-dev.oliver.solutions
sudo git clone <repo-url> /opt/utilisation-dept
cd /opt/utilisation-dept

# Create .env (see "First-time setup" above for value generation).
sudo cp .env.example .env
sudo vi .env

./deploy/deploy.sh

deploy.sh will:

  • Pick a free port in 82008299 and persist it to .env.
  • Render deploy/apache-utilisation-dept.conf from the template.
  • Build the frontend → /var/www/html/utilisation-dept/.
  • Build + start the Docker container.
  • Health-poll /api/health.
  • Print the Apache Include line for the shared vhost.

Wire into the shared vhost (one time only)

Add inside </VirtualHost> of /etc/apache2/sites-enabled/optical-dev.oliver.solutions.conf:

    Include /opt/utilisation-dept/deploy/apache-utilisation-dept.conf

Then:

sudo apachectl configtest && sudo systemctl reload apache2

Smoke-check

curl -sI https://optical-dev.oliver.solutions/utilisation-dept/api/health
curl -sf https://optical-dev.oliver.solutions/utilisation-dept/api/health
# expect 200 and {"ok":true,"version":"0.1.0"}

Then open https://optical-dev.oliver.solutions/utilisation-dept/ in a browser and confirm:

  • Login screen renders.
  • View-source on the loaded JS: search for patHAB — must return nothing.
  • Logging in with the local admin creds shows the dashboard.

Subsequent deploys

ssh user@optical-dev.oliver.solutions
cd /opt/utilisation-dept
./deploy/deploy.sh           # --no-pull, --no-build, --no-frontend, --logs available

Operations

Where the auth log lives

/opt/utilisation-dept/backend/logs/auth.log (mounted into the container at /app/logs/). Rotating, 5 MB × 5 backups.

This is the only audit trail for who logged in when — there's no database.

Rotating the PAT

  1. Generate a new PAT (same scopes as in setup).
  2. Edit .env → update AIRTABLE_PAT.
  3. docker compose -p utilisation-dept restart backend
  4. Revoke the old PAT in the Airtable portal.

Forgot admin password

  1. Generate a new bcrypt hash: python3 -c "from passlib.hash import bcrypt; print(bcrypt.hash('new-password'))"
  2. Edit .env → update ADMIN_PASSWORD_BCRYPT.
  3. docker compose -p utilisation-dept restart backend

Future work

  • Azure AD / MSAL SSO: app/auth/azure.py is stubbed. Flip AUTH_MODE=azure in .env, fill in AZURE_TENANT_ID + AZURE_CLIENT_ID, register the SPA in Azure (platform type: Single-page application), wire MSAL on the frontend, send the ID token to the backend. See ~/.claude/skills/azure-ad-msal-auth.md.
  • Database if/when we need history, audit trails, or to cache more aggressively. Today the data lives in Airtable and the Zoho upload is ephemeral.