Documentation reviewed
Docling
Checked
Recheck due
Docling parses document layout and exports structured document data. Local conversion can run offline after initial model setup; remote-model configurations and managed offerings are separate routes.
- Availability
- Available. Released package/CLI. Docling Serve, remote-model pipelines and IBM's managed offering need separate configuration and terms checks.
- How to use it
- Library / CLI; API. Python 3.10+ and compatible model/runtime dependencies; API serving is a separate deployment step.
- Processing and data
- On-device option; Self-hosted option; Local and cloud options. A local pipeline can process files without uploads after model downloads. Selecting remote enrichment or model providers changes the data destination.
- Costs and limits
- MIT code; hardware, model and optional service costs
- Outputs
- Structured document objects and exports such as Markdown/JSON, tables and provenance; extraction quality must be checked before downstream decisions.
Requires a supported Python environment and downloaded models. Validate tables, reading order, OCR and source locations on your files. The code license does not determine every model's terms.