Sync the full inventory

1. 1

   Build the initial catalog

   Page through /cars, using simple_paginate=1 when totals are unnecessary. Store vehicle ID, source, lot identifiers and update timestamps.

2. 2

   Collect updates

   Poll /cars?minutes=60 for a one-hour window, or pick a window that suits your polling interval (1–4320 minutes). Read every page with the same filters.

3. 3

   Process archived listings

   Poll /archived-lots?minutes=60 over a matching window and mark that source lot inactive. One archived lot does not mean every listing of the vehicle is inactive.

4. 4

   Apply changes idempotently

   Upsert by identifiers. Use overlapping windows and stored timestamps to tolerate retries and duplicates.

5. 5

   Reconcile periodically

   A rolling window is not a durable cursor. After a gap longer than 72 hours, rebuild active inventory instead of trusting the incremental feed.

Archive at lot level using lot_id. A vehicle can have other active lots. Older archive responses contain scalar bid/sale_date; newer responses use {value, updated_at}. Do not assume that a legacy bid is a confirmed final sale price.

A runnable server-side Python example: paginates both feeds, upserts vehicles and lots into SQLite, and archives only the affected lot. Uses a server environment key, timeouts, bounded retries and a last-success checkpoint. Install requests. Set AUCTIONS_API_KEY on the server, then run python sync_auction_cars.py --initial once; schedule subsequent runs without --initial. Default page sizes: cars=50, archive=100. Unlimited can set AUCTIONS_CARS_PER_PAGE=1000 and AUCTIONS_ARCHIVE_PER_PAGE=1000. Run one worker at a time (scheduler lock); adapt SQLite storage and reconcile gaps longer than 72 hours. This example does not make the moving feed a fixed snapshot.