kth-data

KTH data sources

Facts about KTH that recur across tasks, and the one place to look each up. Everything here is a lookup; the editing itself happens with other tools (canvaslms, nytid).

1. Canvas sandbox course: "Sandbox dbosk"

Use the Canvas course "Sandbox dbosk" (Canvas course id 24725) to test any canvaslms command that writes (create/edit/delete pages, assignments, module items, calendar events, quizzes, ...). It is the user's own sandbox; nothing in it is load-bearing.

• Select it with an anchored regex — several other Sandbox courses exist: canvaslms ... -c "^Sandbox dbosk$"
• Prefer reusing existing items; name throwaway artifacts recognizably (e.g. a canvaslms-test-DELETE-ME prefix) and delete them afterwards.
• Verify writes with --no-cache: the CLI updates its local cache on write, so a cached read can mask a server-side no-op.
• Restore whatever you moved or changed when the test is done.
• Never write-test fan-out paths (e.g. create-in-all-courses) against the real server, not even from the sandbox.

Real courses (datintro26, tilkry, ...) are for read-only research (canvas-info skill) unless the task explicitly is to change them.

2. Rooms: https://www.kth.se/places/

KTH's room directory. The page's search box is JavaScript-driven, but https://www.kth.se/places/room/<ROOM NAME> redirects to the room's page (/places/room/id/<uuid>), which lists, among other things:

Label on page	Meaning
Platser:	seats (use this for lecture/seminar capacity)
Examinationsplatser:	seats when used for written exams (much lower)
type line (Hörsal, Övningssal, ...)	room type
Byggnad:, Adress:, Vån:, Campus:	where it is
Utrustning	projector, HDMI, whiteboards, ...

Browsing by building: https://www.kth.se/places/building.

Use the bundled script instead of scraping by hand (TSV: name, seats, exam seats, type, building, address, url; unknown rooms print not found and give exit status 1):

~/.claude/skills/kth-data/scripts/kth_room.py V32 B1 "Ka-Sal C" | column -t -s$'\t'

Room names are the short KTH codes as they appear in schedules (V32, B1, E1, Ka-Sal C, D2, Q1). Adjacent numbering usually means adjacent rooms in the same building (V32/V33/V34/V35, B1/B2/B3) — useful when a large group must be split across rooms. Seat counts are the authoritative capacity; do not reuse numbers from last year's pages without re-checking, rooms get rebuilt.

3. Period and term dates: läsårsindelning

KTH publishes the academic-year calendar (läsårsindelning: terms HT/VT, study periods P1–P4, exam periods, re-exam periods) on the intranet, one page per academic year, e.g. https://intra.kth.se/utbildning/schema-och-lokalbokning/lasarsindelning/lasaret-2026-2027-1.1380361.

Search for the page rather than trusting a remembered URL — the numeric suffix and path change between years and site reorganizations:

• WebSearch for KTH läsårsindelning läsåret <YYYY>-<YYYY+1> (or site:intra.kth.se läsårsindelning <year>); the overview index is https://intra.kth.se/utbildning/schema-och-lokalbokning/lasarsindelning/perioder.
• Then WebFetch the year page and read the dates as written (Swedish YY-MM-DD). Conventions: P1 starts the Monday of the week after "inledande veckor", P1/P2 = HT, P3/P4 = VT; each period ends with a "Tentaperiod"; "Omexaminationsperiod" = re-exams.
• The English version lives under /en/utbildning/... on the same site.

Quote the fetched dates with the source URL in anything you produce; a date from memory is not acceptable for deadlines or schedules.

4. The user's schedule: nytid schedule show

The user's (dbosk's) teaching schedule, including rooms and the TAs booked per event, is available locally:

nytid schedule show --help     # filters: course, date range, register
nytid schedule show            # the user's events, today .. +7 days
nytid schedule show datintro26 --start 2026-08-22 --end 2026-09-01 --no-todo --week

Use it to answer "when/where is X", to find which rooms are booked for an event (e.g. twelve rooms streaming one lecture), and which TAs are booked (Booked TAs: field, sign-up order; --user '.*' does not work). Pair with §2 to get the capacity of every booked room. Details, output columns and gotchas: references/nytid.md.

5. Course register, student numbers, TAs, Zulip (nytid)

Course rounds live in nytid's registers (tcs = the TCS group's shared register on AFS, mine = the user's symlinks into it). A multi-programme course has one merged config plus one subcourse per programme, each with the planning number of students:

nytid courses ls -r tcs | grep datintro26
nytid courses config datintro26-cdate num_students --register tcs
nytid hr users datintro26                 # TA usernames booked on the course
nytid zulip datintro26 streams            # course chat; `send` posts as the user

Use the register's num_students — not Canvas enrolment — when planning rooms before a round starts. How the register, TA roster and Zulip commands work, and what can go wrong: references/nytid.md.

Workflow example: allocate programmes to rooms for a lecture

1. nytid schedule show → find the event and the rooms booked for it.
2. scripts/kth_room.py <rooms…> → seats per room.
3. Course register → expected students per programme.
4. Fit programmes to rooms (keep a split programme in adjacent rooms or at least the same building; B-huset and Q-huset are ~200 m apart, V-huset floors share stairs), leave a margin, and list the mapping as room: seats, PROGRAMME. With little slack a perfect fit is often impossible — prefer a small deficit with "overflow to Zoom" over a programme spread across three floors or a room shared by many programmes. Honour lecturer constraints first (who sits with which programme, who is online).
5. Publish the student-facing mapping in Canvas (canvaslms pages edit; if the course keeps its pages as files in a repo, edit the file and push it from there so the file stays the source of truth). Keep staff-only details (which TA runs which room) out of Canvas: put them in the course Zulip — draft the message for the user to post unless told otherwise.