OCR for construction documents does not work, we fixed it (getanchorgrid.com)

AnchorGrid’s developer documentation describes an API endpoint (POST /v1/drawings/detection/doors) that detects doors in architectural floor-plan PDFs by taking a previously uploaded document_id, enqueuing inference, and returning results via a job/poll workflow. It details which pages can be scanned, how bounding boxes are reported in PDF coordinates after server-side filtering, and the credit-based billing and rate-limiting behavior. The docs also note typical free-tier processing times and that webhook delivery is available only on paid tiers.