Study Room Planner

An independent course prototype (COMP4020 Crit 7) for one slice of ANU Library room booking. Today a group that needs a longer discussion than one room or one person’s allowance covers has to open each library’s booking page separately, compare availability by hand and submit separate reservations. This app puts Chifley, Hancock, Menzies and Law in one view, lets a group assemble a multi-segment plan, and lets a teammate take responsibility for a segment through an in-app invitation. It stores prototype bookings only: availability, rooms and people are fictional demo data, and nothing here reserves a real ANU Library room.

Try it

The home page (/) is a labelled mock-up of today’s ANU Library Bookings page with the planner added as a proposed improvement: the Open Study Room Planner card, the Find a room in every library search box, and the four Book in … Library buttons, which now open the unified planner filtered to that library. The planner itself is at /search.

Cross-room comparison. On the planner page choose Try demo scenario → Cross-room comparison. Four people want 13:00–15:00; no room is free for the whole period, so the best suggestion is Chifley 2.3 then Chifley 2.5. The plan shows the 5-minute room change and says it is 5 minutes short instead of calling it a full match.
Three-hour team plan. Choose Three-hour team plan and add the full match (Chifley 2.3, 13:00–16:00). You own the first two hours; the last hour is over your daily limit, so review the plan, search for Alex Chen’s student number (u9900102), check the name and number in the confirmation dialog and send the invitation. Switch the header’s Demo student to Alex Chen, accept the segment, switch back and Confirm plan. Reload: both reservations are there.
Conflict. Repeat step 2, but as Alex Chen, before accepting, also start a new plan and confirm Chifley 2.3 14:00–15:00 for yourself. The organiser’s confirmation then fails, names the clash and writes nothing: the pending invitation never held the room.
Cancel. On a confirmed plan the organiser can cancel the whole plan, and each booking owner can cancel their own segment. A dialog confirms first; cancelling frees the room and the owner’s daily allowance.

Demo students (fictional): Jordan Lee u9900101, Alex Chen u9900102, Priya Nair u9900103, Tom Walker u9900104, Sofia Rossi u9900105.

Reset demo data in the identity menu clears every plan so the stories can be shown again.

What good looks like here

The source of truth is plan.md (rules, states, wording) and the four reference images/ (visual result). Good means:

Never overstating the plan. Changing rooms costs time (5 minutes within a library, 15 between libraries — design assumptions, not measurements), so a split plan shows the break and its usable discussion time, and a plan short of the request can only be confirmed as an explicit partial plan.
Rules live on the server. Capacity, opening hours, overlap with confirmed bookings and the 2-hour-per-person daily limit (Australia/Sydney calendar day) are rechecked when the plan is confirmed, in one SQLite transaction: every reservation is written, or none is.
Invitations are consent, not holds. An invitation expires 30 minutes after it is sent and never reserves the room; approval only makes the teammate the segment’s booking owner. Nobody transfers quota to anyone.
Honest framing. Letting different group members book consecutive segments for one meeting is a proposed improvement — ANU’s public guidance doesn’t say whether it is allowed.

Enforced by spec/: the planner rules (spec/planner.test.ts); the core flow persisting across reloads, the daily limit, invitation accept/decline/expiry, the no-partial-write conflict case and the page invariants (spec/study-rooms.test.ts, spec/invariants.test.ts). Judgement calls checked by eye: the match with the reference images at 1440px and 390px.

Not built on purpose: ANU SSO, LibCal integration, email, maps, payments, admin tools and a general route optimiser.

Stack

Astro (server output) with Drizzle ORM over SQLite; migrations in drizzle/ apply at boot; deployed to Fly.io with the database on a volume.