Feature: feat/panel to dev

docs/architecture.md

@@ -57,6 +57,8 @@ Edition-scoped event operations. Everything hangs off `hackathon`.
 - Judge roster (`judge`, `judge_assignment`): applications and approval
 - Judging: room-based judging with Devpost import — `judging_*` tables (`judging.ts`). The older pool-dispatch tables (`judging_project`, `judge_vote`, `judge_queue`, `hackathon_result`) are retired and kept for past editions' results
 - Points: `hackathon_event` points, staff adjustments and the point store (`points.ts`)
+- Event check-in rules: nobody is checked into two events that run at the same time, or into two less than 30 minutes apart (meals are exempt). `recordEventAttendance` enforces this for pass scans, event codes and manual check-in
+- Check-in rule-break flags: each refused attempt is recorded (`hackathon_check_in_flag`, `services/check-in-flags.ts`) and shown in judging on the hacker's team, in full to organisers and as a bare marker to judges. Organisers decide whether to pull the team from prizes (`judging_entry.prize_ineligible_at`). A flag outlives the events it names (deleting one clears its id and the organiser view shows "A deleted event"). Hackers never see flags, and they stay out of public and cached payloads
 - Announcements and acceptance waves
 
 Hackathon participation is **open to non-members**. Membership is not a registration requirement.

docs/operations/environment.md

@@ -62,7 +62,7 @@ Without `DATABASE_URL`, `db` is null, sessions fall back to JWT, and tRPC proced
 
 | Variable | Default / notes |
 | --- | --- |
-| `JUDGING_CPSAT_PYTHON` | A Python with OR-Tools (`python3` or a path). Set, the schedule section can run `packages/api/scripts/judging-schedule/cpsat.py` on the server (`judging.schedule.solveCpsat`). Unset means CP-SAT is download, run on a laptop, upload |
+| `JUDGING_CPSAT_PYTHON` | A Python with OR-Tools (`python3` or a path). Set, the schedule section can run `packages/api/scripts/judging-schedule/cpsat.py` on the server (`judging.schedule.solveCpsat`), and `judging.schedule.assignJudges` uses it to pick judges. Unset means CP-SAT is download, run on a laptop, upload, and the app assigns judges itself |
 | `JUDGING_CPSAT_SCRIPT` | Path to `cpsat.py`. Default: found by walking up from the working directory |
 
 ## Security / proxy

docs/operations/testing.md

@@ -28,6 +28,7 @@ Config: `packages/api/vitest.config.ts`. Includes `src/**/*.test.ts` and `src/.i
 | Hackathon flow | `hackathon-flow.test.ts`, `hackathon-admin-edge.test.ts`, `hackathon-interest.test.ts` |
 | Payments | `stripe-payments.test.ts` |
 | QR / check-in | `qr-checkin.test.ts` |
+| Check-in rule-break flags | `services/check-in-flags.test.ts`, `.internal-tests/check-in-flags-*.test.ts` |
 | Security / sanitizer / rate limit | `security.test.ts`, `resilience.test.ts` |
 | Judges / participants / initiatives / bootcamp / announcements | matching `*-edge.test.ts` / `bootcamp.test.ts` / `announcements.test.ts` |
 | Portal context | `services/portal-context.test.ts`, `routers/user/portal-context.test.ts` |

docs/packages/api.md

@@ -91,7 +91,13 @@ Batch status updates allow up to **2500** ids. The global array cap in `scrubMar
 
 **Admin** (`admin.ts`): `list`, `create` (add by email), `setActive` (approve or suspend), `remove` (super admin).
 
-Judging itself is the `judging` namespace; its shared helpers are in `services/judging/` (`room-link`, `scope`, `scoring`, `config`). `hackathon.getResults` returns published placings from both the retired `hackathon_result` table and `judging_placement`, in one shape.
+Judging itself is the `judging` namespace; its shared helpers are in `services/judging/` (`room-link`, `scope`, `scoring`, `config`). `hackathon.getResults` returns published placings from both the retired `hackathon_result` table and `judging_placement`, in one shape. It leaves out entries pulled from judging (`deleted_at`) or from prizes (`prize_ineligible_at`, set by `judging.admin.setPrizeEligibility`). Check-in rule-break flags (`services/check-in-flags.ts`) reach organisers through `judging.admin.listEntries` and `judging.results.list`, and judges through `judging.desk` as a bare `flagged` boolean; no hacker-facing, public or cached procedure returns them.
+
+Placings rank on scores adjusted for judge bias (`services/judging/calibration.ts`): each evaluation is calibrated within its root challenge for its judge's strictness, spread and drift, on the assumption that `judging.desk.nextProject` (`services/judging/queue.ts`) hands every judge a random sample of projects. `judging_placement.raw_score` keeps the plain mean; only `judging.results.list` (organisers) returns it and the per-judge calibration, and `judging.desk` stays on raw means. An evaluation counts when it holds a rating for every rating item (`scoring.hasEveryRating`, and `scoring-sql.holdsEveryRating` where the database counts), whether or not a written response is missing, so a draft submitted at the hard end still places its project. Organiser counts, the board and hacker itinerary status, hacker feedback, and the desk's scores, judged count and shortlist all use that rule. `judging.results.list` keeps the calibration for 60 seconds per edition, keyed on the edition's evaluation count and latest save and on its ranked entries, while `judging.results.compute` always calibrates fresh. `judging.admin.reopenGrade` deletes one evaluation and the drafts for its project and challenge so another judge can grade it, and clears the edition's unpublished placings, so publishing needs a fresh compute; it is refused while results are published.
+
+At an untimed challenge that random sample is enforced (`services/judging/queue-store.ts`): `judging.desk.nextProject` offers only the first `QUEUE_WINDOW` (3) open projects of the judge's own queue and takes a `skip` list for a team away from its table, and a judge's first `startGrading`, `saveDraft` or `submit` on a project outside that window is refused. `judging.desk.startGrading` holds a project from the moment it is opened, with an empty draft. A table judge grades a sponsor's challenge only from a seat in a room that serves it (`scope.isSponsorChallenge`); `judging.desk.context` lists only the challenges they may pick.
+
+A scheduled presentation can have one assigned judge (`judging_appointment.assigned_judge_id`, set by `judging.schedule`; `services/judging/schedule/judges.ts` decides who may be assigned it). The desk never writes it and never tells a judge which other judge is assigned to, holds or graded a project. From the presentation's start until `TAKEOVER_AFTER_MS` (5 minutes) later only the assigned judge may open it (`services/judging/timing.ts`), and other judges see it as `reserved` with the time it opens. After that any judge who may grade it there can take it over, unless the assigned judge has started: their draft holds it (`grading-hold.ts`). Every claim checks this under the grading lock. `judging.desk.nextProject` offers the judge's own running assignments first; the queue window does not apply to scheduled challenges. `judging.desk.itinerary` lists the caller's assigned presentations that have not ended, soonest first, with the room and whether they are seated there. `judging.desk.releaseGrading` deletes the caller's own draft while it is still the empty one `startGrading` made, so a skipped or accidentally opened project is free again. `judging.schedule.get` gives organisers each started assignment's `judgeProgress` (grading, graded, taken over, not started).
 
 ## Club procedures (selected)
 

docs/packages/db.md

@@ -27,7 +27,7 @@ Files in `src/schemas/`, re-exported from `schemas/index.ts`. `drizzle.config.ts
 | `auth.ts` | `user`, `account`, `session`, `verificationToken` |
 | `members.ts` | `user_profile`, `member`, `membership_history` |
 | `admins.ts` | `admin` |
-| `hackathons.ts` | `hackathon`, `hackathon_team`, `hackathon_participant`, `hackathon_project`, `hackathon_interest`, `hackathon_event`, `hackathon_event_attendee`, `hackathon_announcement`, `hackathon_announcement_recipient` |
+| `hackathons.ts` | `hackathon`, `hackathon_team`, `hackathon_participant`, `hackathon_project`, `hackathon_interest`, `hackathon_event`, `hackathon_event_attendee`, `hackathon_check_in_flag`, `hackathon_announcement`, `hackathon_announcement_recipient` |
 | `judge.ts` | `judge`, `judge_assignment` (live roster); `judging_project`, `judge_vote`, `judge_queue`, `hackathon_result` (retired, kept for past editions) |
 | `judging.ts` | `judging_*`: config, entries and members, challenges, buildings, rooms and room links, judges, guest sessions, presence, schedule and appointments, rubric, evaluations, deliberation, placements |
 | `points.ts` | `hackathon_point_adjustment`, `hackathon_store_item`, `hackathon_store_purchase` |