docling-project/docling-serve

Docling Serve is an API service that exposes Docling document processing capabilities over HTTP.

The tool solves the problem of integrating document conversion and analysis into applications by wrapping Docling as a web service. It provides REST endpoints that accept documents and return structured extraction results, eliminating the need to embed Docling directly in application code. The service runs on FastAPI and includes interactive API documentation and a UI playground for testing conversions.

Developers should choose this tool when they need to process documents through a network service rather than as a library dependency, or when they want to scale document processing independently from their main application. It suits projects that require document conversion as a microservice, containerized deployments, or multi-tenant setups where document processing should be isolated. The tool offers multiple container images optimized for different hardware configurations: a base image supporting both CPU and GPU, a CPU-only variant, and CUDA-specific builds for GPU acceleration on different versions.

The project maintains a stable v1 API with migration documentation for users upgrading from earlier versions. Development includes comprehensive documentation covering server configuration, runtime options, and deployment examples. The tool is distributed through multiple container registries with support for both x86 and ARM architectures, indicating attention to deployment flexibility across different infrastructure environments.