CR-004 and CR-018 are the same change seen from two ends. CR-018 is why the
Acumatica cost code came out rather than being relabelled — the tracking
dimension the team wants is floor and area, not an accounting code — and CR-004
is what makes that dimension exist. Committed together because a rollup keyed on
free text is not a rollup, and structured location with nothing rolling up by it
is a form change nobody asked for.
CR-004 - three dependent dropdowns
Building filters Floor filters Sector, off the project's own taxonomy from
T5.4. Clearing a parent clears its children: not doing that is how a package
ends up filed under a floor that is not in the building it claims.
PATHS are stored, not names and not bare codes. A floor's own code is not
unique across buildings; `B-ONE/L1` is. That is what lets the dashboard filter
by a building and match everything beneath it with a prefix test, and it is
what CR-018 groups on.
The list is fetched with include_inactive=true, which is not a contradiction of
CR-005's "deactivating hides it from new work packages" — they are two
questions. What may be CHOSEN is active only. What may be SHOWN is everything,
because a package already referencing a deactivated value still has to render
its label, and blanking it on open would write the blank back on the next save.
A deactivated value that IS on the package is offered, labelled "(no longer
offered)"; on a fresh package it is not offered at all. Both checked.
X5, checked the way aggregates_check checks its own: localStorage is poisoned
with a fake building and the dropdown is required to ignore it.
wp_location survives as a hidden field. A package written before this keeps
what it said, and the form says so rather than dropping it.
CR-018 - the rollup
LOCATION_DIMENSIONS is now ("building", "floor", "sector"). T4.1's note said
"only this tuple and the keys inside each group change - the response shape
does not", and that held exactly.
Rolled up at EVERY level, server-side, not just at the leaf. "How many on
floor 2" is the question CR-018 asks and it is a level above the leaf groups;
summing them in the browser would be the same per-browser arithmetic B4
removed. Actual Hours rolls up along the same dimensions - that is the field
CR-017 retained, and this is why that decision mattered.
Packages with no location are an explicit "(unassigned)" row, not a gap. The
reason is arithmetic: a group set that silently omits them does not add up to
the project total, and a rollup that does not reconcile is decoration. The
probe checks every level sums to the project total, and to the estimated and
actual hour totals, using distinct primes so a mis-sum cannot land on the
right number by luck.
A package with a building but no floor lands in the floor-level unassigned row
alongside the one with no location at all - which is the honest answer, and is
asserted by its hours rather than by its count.
Free text captured before CR-004 groups under itself as a building rather than
collapsing into unassigned, one level deep. Pretending free text is a
hierarchy would file "FAB / LVL 1" under a building called "FAB / LVL 1".
server/app.py dimensions, _location_levels, hours per group
html/wp-creation-index.html three selects where the text box was
html/wp-creation-app.js the pickers, the filters, the rollup panel
html/wp-creation-styles.css .loc-picker, .loc-rollup
tests/rollup_check.py new - 63 checks
Done when — CR-004
[x] all three render as dropdowns populated from project configuration
[x] dependent filtering works, and clearing a parent clears its children
[x] values persist as codes; confirmed by reading what collectPackage stored
[x] the dashboard filters by each of the three
[x] a work package referencing a deactivated value still renders correctly
[x] all option data comes from the server - proved by poisoning the cache
Done when — CR-018
[x] the dashboard groups and totals by Building, Floor and Sector
[x] totals reconcile against an unfiltered count, at every level
[x] Actual Hours rolls up along the same dimensions
[x] grouping is computed server-side - proved by putting nine fake packages in
localStorage and requiring the panel to show none of them
[x] work packages with no location appear in an explicit unassigned group
No migration: location lives in the work package's JSON data blob like every
other per-package field. No colour literal added.
Verified one at a time
rollup_check 63/63 new
generalinfo_check 49/49
browser_check 71/71
pipeline 43/43
a11y 22/22
aggregates 16/16
f_items F1-F5 FIXED, F6 REPRODUCES (T7.2)
Question for the PR, per CLAUDE.md: the dashboard's location filters and the
rollup both key on the path, so a package saved with free text and no codes is
unreachable by any location filter and sits in its own building-level row. That
is correct and it is also a migration question - whether the existing free-text
locations should be mapped onto the taxonomy once the B100 list arrives, or left
as history. Nothing here decides it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Work Package Suite API
A small Python (FastAPI) service that stores project SOPs, Work Packages,
and comments in PostgreSQL. NGINX serves the static site and proxies /api/
to this service.
browser → NGINX ──serves──> static site (index.html, …)
└─proxy /api/─> api container (:8000) → db container (postgres)
Endpoints
| Method | Path | Purpose |
|---|---|---|
| GET | /api/health |
liveness check (unauthenticated) |
| POST | /api/auth/login |
sign in ({username, password}) — sets the session cookie |
| POST | /api/auth/logout |
clear the session cookie |
| GET | /api/auth/me |
the logged-in user |
| POST | /api/auth/password |
change your own password |
| GET | /api/auth/users |
list accounts (admin) |
| POST | /api/auth/users |
create an account (admin) |
| DELETE | /api/auth/users/{id} |
delete an account (admin) |
| POST | /api/sops |
create/update a SOP (upsert by id) |
| GET | /api/sops |
list SOP summaries |
| GET | /api/sops/latest?complete=true |
most recent (complete) SOP |
| GET | /api/sops/{id} |
full SOP document |
| DELETE | /api/sops/{id} |
delete a SOP |
| POST | /api/wps |
create/update a Work Package (upsert by id) |
| GET | /api/wps?sop_id=… |
list WPs (optionally for one SOP) |
| GET | /api/wps/{id} |
full WP document |
| DELETE | /api/wps/{id} |
delete a WP |
| POST | /api/comments (and /api/feedback) |
add a comment |
| GET | /api/comments?source=&sop_id=&wp_id=&step= |
list comments |
Interactive docs once running: /api/docs.
The full client document is stored in each row's data (JSON) column; common
fields (name, number, status, …) are promoted to columns for listing/filtering.
Login portal (user accounts)
The suite is gated by a username/password login. Sign-in issues a signed JWT
that rides in an HttpOnly, SameSite=Lax cookie (wp_session); the cookie is
marked Secure automatically whenever the request arrives over HTTPS (via
NGINX's X-Forwarded-Proto). There is no server-side session store — each
request is validated by checking the cookie's signature and expiry.
The real security boundary is the API: every /api/ data route is refused
with 401 unless a valid session cookie is present (see auth_gate in
app.py). The static pages additionally include auth-guard.js, which redirects
to login.html when there's no session — that's for UX, not protection.
Passwords are stored only as bcrypt hashes (server/auth.py). Roles are
admin (may manage users) and user.
Set the signing secret
Add AUTH_SECRET_KEY to .env (see .env.example). Required in production —
without it the API uses a random per-process key, so logins reset on restart.
python -c "import secrets; print(secrets.token_urlsafe(48))"
Create the first admin
The /api/auth/users endpoint needs an existing admin, so bootstrap one from a
shell (run from the project root, like uvicorn):
python -m server.manage_users create-admin alice --name "Alice Smith"
# prompts for a password (min 8 chars)
In Docker:
docker compose exec api python -m server.manage_users create-admin alice --name "Alice Smith"
Other commands: create <user> --role user, list, reset-password <user>,
disable <user>, enable <user>. After that, admins can add users through the
API (or you can keep using the CLI).
Local dev
cd server
python -m venv .venv && . .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -r requirements.txt
# No DATABASE_URL → uses a local sqlite file, so you can start immediately:
uvicorn server.app:app --reload --port 8000 # run from the PROJECT ROOT
Then open http://localhost:8000/api/docs.
Run uvicorn/gunicorn from the project root (the folder that contains the
server/directory), because the import path isserver.app:app.
Production — Docker Compose
This is the recommended production setup. Three containers run in an isolated
internal network; only NGINX is exposed to the outside via the external proxy
network.
[external proxy network]
│
┌────▼────┐ internal network ┌──────────┐ ┌────────┐
│ nginx │ ───────────────────> │ api │ → │ db │
└─────────┘ └──────────┘ └────────┘
1. Create the credentials file
Create .env in the project root (same directory as docker-compose.yml).
This file is never committed — add it to .gitignore.
# .env — project root
POSTGRES_DB=wpsuite
POSTGRES_USER=wpsuite
POSTGRES_PASSWORD=<strong-random-password>
# Must match POSTGRES_* above; hostname is the compose service name "db"
DATABASE_URL=postgresql+psycopg://wpsuite:<strong-random-password>@db:5432/wpsuite
Generate a strong password:
openssl rand -base64 32
2. Add the Dockerfile
Create Dockerfile in the project root:
FROM python:3.12-slim
WORKDIR /app
COPY server/requirements.txt ./server/
RUN pip install --no-cache-dir -r server/requirements.txt
COPY server/ ./server/
EXPOSE 8000
CMD ["gunicorn", "-k", "uvicorn.workers.UvicornWorker", \
"-b", "0.0.0.0:8000", "--workers", "2", "server.app:app"]
3. Update the NGINX site config
The API is no longer at 127.0.0.1:8000 — it is the api container.
Update the /api/ proxy block in your nginx conf (e.g. nginx/conf.d/wp-suite.conf):
location /api/ {
proxy_pass http://api:8000; # ← service name, not localhost
proxy_http_version 1.1;
proxy_set_header Host $host;
proxy_set_header X-Forwarded-For $remote_addr;
proxy_set_header X-Forwarded-Proto $scheme;
client_max_body_size 5m;
}
4. docker-compose.yml
Replace your existing docker-compose.yml with:
services:
webserver:
image: nginx:alpine
container_name: nginx_webserver
volumes:
- ./html:/usr/share/nginx/html:ro
- ./nginx/conf.d:/etc/nginx/conf.d:ro
- ./nginx/nginx.conf:/etc/nginx/nginx.conf:ro
- ./logs:/var/log/nginx
restart: unless-stopped
depends_on:
api:
condition: service_started
networks:
- proxy # external — reachable by your reverse proxy / traefik
- internal # needs a path to the api container
api:
build: .
container_name: wp_api
env_file: .env # loads DATABASE_URL
restart: unless-stopped
depends_on:
db:
condition: service_healthy # waits for postgres to accept connections
networks:
- internal
db:
image: postgres:16-alpine
container_name: wp_db
env_file: .env # loads POSTGRES_DB / USER / PASSWORD
volumes:
- pgdata:/var/lib/postgresql/data
restart: unless-stopped
healthcheck:
test: ["CMD-SHELL", "pg_isready -U $${POSTGRES_USER} -d $${POSTGRES_DB}"]
interval: 10s
timeout: 5s
retries: 5
networks:
- internal
volumes:
pgdata:
networks:
proxy:
name: proxy
external: true
internal:
internal: true # no outbound internet access from api/db
5. First-time startup
# Build the api image and start all containers
docker compose up -d --build
# Confirm all three containers are running
docker compose ps
# Tail logs (Ctrl-C to stop following)
docker compose logs -f api
Tables are created automatically on first API startup — no manual CREATE TABLE
needed.
Authentication notes
Postgres → API authentication is handled entirely through DATABASE_URL in
.env. The db container uses POSTGRES_USER / POSTGRES_PASSWORD to
initialise the database on first run; the api container uses the matching
credentials in DATABASE_URL to connect. Neither credential ever appears in the
compose file itself.
Network isolation: the db container is on the internal network only —
it has no port exposed to the host and is unreachable from outside the compose
stack. Only the api container can open a connection to it.
Changing the password: update both POSTGRES_PASSWORD and the password
in DATABASE_URL in .env, then:
# Stop api first (db must keep running to accept the ALTER USER command)
docker compose stop api
docker compose exec db psql -U wpsuite -c "ALTER USER wpsuite PASSWORD 'new-password';"
docker compose start api
Day-to-day operations
# Rebuild api after a code change
docker compose up -d --build api
# View postgres data directly
docker compose exec db psql -U wpsuite -d wpsuite
# Take a database backup
docker compose exec db pg_dump -U wpsuite wpsuite > backup-$(date +%F).sql
# Restore from backup
docker compose exec -T db psql -U wpsuite -d wpsuite < backup-2025-01-01.sql
# Stop everything (data volume is preserved)
docker compose down
# Stop everything AND delete all data
docker compose down -v
Quick test
/api/health is open; data routes now require a session, so log in first and
reuse the cookie jar:
curl http://127.0.0.1:8000/api/health # {"ok":true} — no auth needed
# Sign in, saving the session cookie to a jar
curl -c jar.txt -X POST http://127.0.0.1:8000/api/auth/login \
-H 'Content-Type: application/json' \
-d '{"username":"alice","password":"<password>"}'
# Reuse the cookie on protected routes
curl -b jar.txt http://127.0.0.1:8000/api/comments
Without the cookie, protected routes return 401 {"detail":"Not authenticated"}.
Or via the nginx proxy (replace with your hostname):
curl https://wp-suite.company.local/api/health