Set up the VM: PostgreSQL, Node, data import and backend service

[ivankoc]

Goal

Get the database and backend running on the group VM, directly on the OS (no Docker), so the prototype can talk to it.

Acceptance criteria

• Confirm the VM hostname (assumed it2810-13.idi.ntnu.no) and update CLAUDE.md if different
• sudo apt install postgresql unzip, role for the deploy user (sudo -u postgres createuser --createdb $USER), createdb t13
• Node.js >= 22.18 installed (NodeSource or nvm)
• Repo cloned, npm ci, npm run import runs on the VM (downloads data itself, no copying needed)
• Import result matches the reference run: 17,698 recipes, 5,653 with photo
• Backend runs as a systemd service on port 3001, restarts on reboot/crash
• Apache serves data/images/ (e.g. /project2/images/) with long cache headers
• All steps documented in README ("Deployment" section) so staff can restart everything

Protect against dataset changes

The import downloads both datasets from Kaggle at run time. If a dataset is changed or removed there, we can no longer rebuild the same database. Two safeguards:

• Keep a copy of the raw zips (data/raw/*.zip, ~228 MB) outside Kaggle, e.g. on the VM in a backup folder outside the repo, or on the team's shared drive. Document where in the README.
• Import checks its own output: record the SHA-256 of both zips and the expected counts (17,698 recipes / 5,653 photos) in import.ts. Fail with a clear message if a download's checksum differs or the counts don't match, so a changed dataset is noticed immediately instead of silently giving different data.

Notes

Depends on #12 (data import) being merged. How to start the backend must be documented on git or in a file on the VM (assignment requirement).

[davidhne]

Status before the next session.

Done

• Hostname confirmed: it2810-13.idi.ntnu.no (matches CLAUDE.md)
• PostgreSQL 18 and unzip are already installed on the VM, and Postgres is running
• /var/www/html/project2/ exists and serves the current frontend build: http://it2810-13.idi.ntnu.no/project2/
• The GraphQL API is merged (feat(api): recipe schema and resolvers backed by Postgres #22), so there is a real backend to run

Not done yet

• Postgres role for the deploy user and createdb t13 (role davidhne does not exist yet)
• Node.js on the VM (not installed)
• Repo on the VM, npm ci, npm run import. Note: the VM has no read access to git.ntnu.no yet, so cloning needs a deploy key (or copy the repo over with rsync)
• Backend as a systemd service on port 3001
• Apache serving data/images/
• README "Deployment" section

Two things to add to this issue

Both are small changes in the Apache site config and need sudo.

1. Proxy the API through Apache. The frontend runs in the visitor's browser, so the browser has to reach the backend. Port 3001 is not reachable from outside today, and relying on it is fragile. Forward /project2/graphql to localhost:3001 instead, and point the frontend at /project2/graphql. Same origin, so no CORS either.
2. SPA fallback for React Router. /project2/fridge returns 404 on reload or direct link, because Apache looks for a file called fridge. .htaccess is ignored on this VM (AllowOverride None): /project1/favorites also returns 404, so project 1's .htaccess never took effect. Put the rule in the Apache config instead.

# sudo a2enmod proxy proxy_http
<Location /project2/graphql>
  ProxyPass http://localhost:3001/
  ProxyPassReverse http://localhost:3001/
</Location>

<Directory /var/www/html/project2>
  FallbackResource /project2/index.html
</Directory>

Then sudo apachectl configtest && sudo systemctl reload apache2.