# Docling (/docs/coreai/components/docling)



Component Category [#component-category]

Document AI inference

Component Description [#component-description]

Docling is the document processing component used to parse documents into structured representations that can be consumed by downstream AI workflows.

Why It Is Used [#why-it-is-used]

It helps CoreAI turn raw documents into usable content for retrieval, knowledge workflows, and document intelligence features.

Learn More [#learn-more]

* [Docling Documentation](https://docling-project.github.io/docling/)
* [Docling on GitHub](https://github.com/docling-project/docling)

Developer Guidance [#developer-guidance]

For application integrations on BullSequana AI, use the `CoreAI API` upload flow (`/v1/files/upload`) instead of calling Docling directly. This keeps integrations aligned with the platform contract and allows backend workflow orchestration and resource controls to evolve without breaking clients.

GPU acceleration [#gpu-acceleration]

Docling runs in CPU mode by default. GPU acceleration can be enabled through an environment-variable override, which switches the image to the CUDA build and adds `nvidia.com/gpu` resource requests. See [Enable GPU for AI components](/docs/deployment/playbooks/enable-gpu).

Interacts With [#interacts-with]

* `CoreAI API` for document ingestion and processing flows
* `Milvus` when parsed content is embedded and stored for retrieval
* `Rook Ceph` for document and artifact storage through RGW
* `LiteLLM` and other CoreAI workflows where processed content is used downstream