Using the API
Pagination
Collections are paged in one of two ways, and each operation's reference says which.
Cursor pages (for syncing)
GET /members, GET /loans and GET /holds use a cursor, because their data changes while you read it.
GET /lms/v1/members?updated_since=2026-11-01T00:00:00Z&limit=100 HTTP/1.1
Host: api.librarika.dev.raynux.com
Authorization: Bearer $LIBRARIKA_ACCESS_TOKEN
{"items": [{"id": 1001, "member_no": "A-0001", "status": "active", "updated_at": "2026-11-02T09:30:00Z"}], "next_cursor": "eyJ2IjoxfQ"}
limitis 1 to 100 (default 20).- Pass
next_cursorback ascursor, with the same filters, untilnext_cursorisnull. - Rows come in a stable order (
updated_at, thenid), so a sync never misses one. Start a sync withupdated_sinceand keep the highestupdated_atyou saw for the next run. - A cursor is opaque, bound to your client and to its filters, and expires after 24 hours. A changed, expired or foreign cursor is
400 invalid_cursor: start again from your lastupdated_since.
Numbered pages (for search)
GET /catalog/titles, GET /me/loans and GET /me/holds use page numbers:
{"items": [], "page": 1, "limit": 20, "total": 0, "page_count": 0}
pagestarts at 1;limitis 1 to 100 (default 20).- A search reaches the first 1,000 results only (
page×limit≤ 1,000); deeper is400 page_too_deep. Narrow the search instead.
Small lists
GET /library/branches answers all branches as {"items": [...]}, without paging. GET /catalog/titles/availability?ids= takes 1 to 50 ids and answers {"items": [...]}; ids not in the library are left out.