Docling support for AI Accelerators

[nikos-livathinos]

🚀 Get ready for the next update! As of Docling 2.12.0, there is support for AI accelerators (AUTO, CPU, CUDA, MPS). The device, num_threads API/CLI parameters and their corresponding envvars DOCLING_DEVICE, DOCLING_NUM_THREADS allow to control the hardware settings used to run Docling. The AUTO option automatically selects the best device available in the system.

👉 This example demonstrates how to use Docling with accelerators.