feat: added judge frontend and backend #2

Open
angelayu wants to merge 4 commits from feature/judge_backend_and_frontend into main
Member

Summary

Adds the first working version of the Nova judge tool from RFC 0001 (0001-nova-summarization-tool.md) with Google sign-in, so judges can sign in, browse hackathon submissions by subteam, and take notes on each project.

Judges sign in with Google. The backend verifies the Google token, checks the email against a judge list loaded from a CSV (RFC section 6), and starts a session. Private notes are now limited to the signed-in judge's subteam on the server. The RFC lists Better Auth, but it is a JavaScript library for a Node server, so this verifies Google tokens directly in the FastAPI backend instead.

AI summaries, /ask and the rest of the RFC are not included and will come in follow-up PRs.

Changes

  • Add nova-backend, a FastAPI app that reads the submissions Google Sheet (published as CSV) and returns one project per team, with a blank subteam shown as "Unassigned".
  • Add SQLite-backed notes with endpoints to list and add notes per project, including a "share with all judges" option.
  • Add Google sign-in: POST /auth/google, GET /auth/me and POST /auth/logout, with a signed session in an httpOnly cookie.
  • Require sign-in on /projects and the notes endpoints. Note author and subteam now come from the signed-in user, not the request, and admins see all notes.
  • Add load_users.py to load judge and admin accounts from a CSV (judges.example.csv shows the format).
  • Load config from nova-backend/.env (see env.example): SHEET_CSV_URL, GOOGLE_CLIENT_ID and SESSION_SECRET.
  • Add nova-judge, a React + TypeScript (Vite) frontend with a Google sign-in screen, a sidebar of teams grouped by subteam, search, and a project page with the GitHub and slides links.
  • Add note-taking with an optional feature tag, and a summary panel that refreshes every 7 seconds.
  • Add START-HERE.md with local setup steps (will be deleted once README.md is created).

Not included yet

  • Emailed magic-link sign-in fallback
  • AI-generated summaries and /ask
  • GitHub README and slide enrichment, the video pipeline, and the admin page

How to test

  1. Create a Google OAuth web client with http://localhost:5173 as an authorized origin.
  2. In nova-backend, copy env.example to .env, fill in the three values, then run python load_users.py judges.csv with your own email in the CSV.
  3. Start the backend with uvicorn main:app --reload --port 8000 and the frontend with npm run dev (with VITE_GOOGLE_CLIENT_ID in nova-judge/.env.local).
  4. Open http://localhost:5173 and sign in with a listed account. An unlisted account should be rejected.
  5. Add a private and a shared note, then sign in as a judge from another subteam and confirm only the shared note is visible.

AI use

  • None
  • Minor: autocomplete or suggestions for a few lines
  • Some: AI drafted parts of the change, the rest was written by hand
  • [√] Most: AI wrote the bulk of the change, edited by hand
  • All: AI wrote the whole change, reviewed and tested by hand

Checklist

  • [√] Backend starts, sign-in works, and /projects loads from a published sheet
  • [√] npm run lint has no errors (3 set-state-in-effect warnings) and npm run build passes in nova-judge
  • README updated if a variable, route, or command changed
  • [√] No secrets, keys, or .env values in the diff
  • [√] Commits follow Conventional Commits, one change each, rebased onto main
  • [√] Title written like a commit subject
## Summary Adds the first working version of the Nova judge tool from RFC 0001 (`0001-nova-summarization-tool.md`) with Google sign-in, so judges can sign in, browse hackathon submissions by subteam, and take notes on each project. Judges sign in with Google. The backend verifies the Google token, checks the email against a judge list loaded from a CSV (RFC section 6), and starts a session. Private notes are now limited to the signed-in judge's subteam on the server. The RFC lists Better Auth, but it is a JavaScript library for a Node server, so this verifies Google tokens directly in the FastAPI backend instead. AI summaries, `/ask` and the rest of the RFC are not included and will come in follow-up PRs. ## Changes - Add `nova-backend`, a FastAPI app that reads the submissions Google Sheet (published as CSV) and returns one project per team, with a blank subteam shown as "Unassigned". - Add SQLite-backed notes with endpoints to list and add notes per project, including a "share with all judges" option. - Add Google sign-in: `POST /auth/google`, `GET /auth/me` and `POST /auth/logout`, with a signed session in an httpOnly cookie. - Require sign-in on `/projects` and the notes endpoints. Note author and subteam now come from the signed-in user, not the request, and admins see all notes. - Add `load_users.py` to load judge and admin accounts from a CSV (`judges.example.csv` shows the format). - Load config from `nova-backend/.env` (see `env.example`): `SHEET_CSV_URL`, `GOOGLE_CLIENT_ID` and `SESSION_SECRET`. - Add `nova-judge`, a React + TypeScript (Vite) frontend with a Google sign-in screen, a sidebar of teams grouped by subteam, search, and a project page with the GitHub and slides links. - Add note-taking with an optional feature tag, and a summary panel that refreshes every 7 seconds. - Add `START-HERE.md` with local setup steps (will be deleted once README.md is created). ## Not included yet - Emailed magic-link sign-in fallback - AI-generated summaries and `/ask` - GitHub README and slide enrichment, the video pipeline, and the admin page ## How to test 1. Create a Google OAuth web client with `http://localhost:5173` as an authorized origin. 2. In `nova-backend`, copy `env.example` to `.env`, fill in the three values, then run `python load_users.py judges.csv` with your own email in the CSV. 3. Start the backend with `uvicorn main:app --reload --port 8000` and the frontend with `npm run dev` (with `VITE_GOOGLE_CLIENT_ID` in `nova-judge/.env.local`). 4. Open http://localhost:5173 and sign in with a listed account. An unlisted account should be rejected. 5. Add a private and a shared note, then sign in as a judge from another subteam and confirm only the shared note is visible. ## AI use - [ ] None - [ ] Minor: autocomplete or suggestions for a few lines - [ ] Some: AI drafted parts of the change, the rest was written by hand - [√] Most: AI wrote the bulk of the change, edited by hand - [ ] All: AI wrote the whole change, reviewed and tested by hand ## Checklist - [√] Backend starts, sign-in works, and `/projects` loads from a published sheet - [√] `npm run lint` has no errors (3 `set-state-in-effect` warnings) and `npm run build` passes in `nova-judge` - [ ] README updated if a variable, route, or command changed - [√] No secrets, keys, or `.env` values in the diff - [√] Commits follow Conventional Commits, one change each, rebased onto `main` - [√] Title written like a commit subject
feat: added judge frontend and backend
All checks were successful
kennel/build skipped: no services or static sites defined
3ab91b8a77
feat: added google authentication for judges
All checks were successful
kennel/build skipped: no services or static sites defined
1d1e0b79e5
Member

Thanks for all the work on this so far! Just a few notes:

Please include a description for each PR summarizing the changes made and for some examples, you can look at some of the merged Bark Agent or Surface PRs.

Also, we should remove node_modules from Git's tracking index. I believe it's already in .gitignore, but it was likely committed before the ignore rule was added, which is why the PR is showing around 1 million lines of changes.

Finally, since the two PRs are related, I think it would make more sense to consolidate them into a single PR. There's no need to have two separate ones for now.

Thanks for all the work on this so far! Just a few notes: Please include a description for each PR summarizing the changes made and for some examples, you can look at some of the merged Bark Agent or Surface PRs. Also, we should remove node_modules from Git's tracking index. I believe it's already in .gitignore, but it was likely committed before the ignore rule was added, which is why the PR is showing around 1 million lines of changes. Finally, since the two PRs are related, I think it would make more sense to consolidate them into a single PR. There's no need to have two separate ones for now.
fix: removed node_modules from tracking
All checks were successful
kennel/build skipped: no services or static sites defined
93af31e142
fix: removed notes.db from tracking
All checks were successful
kennel/build skipped: no services or static sites defined
fef7dea98b
Some checks are pending
kennel/build skipped: no services or static sites defined
nixbot/nix-eval
Required
CI / checks*
Required
kennel/deploy
Required
This pull request doesn't have enough approvals yet. 0 of 1 approvals granted.
You are not authorized to merge this pull request.
View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u origin 'feature/judge_backend_and_frontend:feature/judge_backend_and_frontend'
git switch 'feature/judge_backend_and_frontend'
Sign in to join this conversation.
No description provided.