Roboflow Inference Server API
The OpenAPI-described inference surface of the Roboflow platform. Runs trained Roboflow models plus a catalog of foundation models — CLIP and Perception Encoder embeddings, Grounding DINO, YOLO-World and OWLv2 open-vocabulary detection, SAM/SAM2/SAM3 segmentation, DocTR/EasyOCR/TrOCR/PP-OCR text extraction, depth estimation, and large multimodal models — over HTTP. The same specification is served by the hosted serverless endpoints and by a self-hosted Docker inference server.