Self-Hosted (On-Premises)
Run DocAI Fabric on your own infrastructure using Docker. You keep complete control of the hosting environment and your data never leaves it, apart from the calls to whichever language models you choose to connect and to the licensing service.
DocAI Fabric is packaged as a single Docker container that includes the API server, the workflow workers, the built-in OCR engines (PaddleOCR and Tesseract), and the web interface. The only supporting service it needs is Redis for the job queue. Because OCR runs inside the container, no cloud OCR service is required.
Choose a model
| Model | Best for | Nodes | Survives a node failure |
|---|---|---|---|
| Standalone (Single Node) | Demos, evaluations, small workloads | 1 | No |
| High Availability (Multi-Node) | Production | 2 or more | Yes |
Both models use the same application image and the same built-in OCR. The difference is redundancy and scale: Standalone runs everything on one machine (a virtual machine or even a laptop), while High Availability spreads the application across several nodes that share storage and a Redis queue behind a load balancer.
Start with Standalone (Single Node) to get going quickly, then move to High Availability when you are ready for production.
What is the same in both
- One image, no external OCR. Classification and extraction use the language model endpoints you connect on the Models page; OCR runs locally inside the container.
- You connect the models in the app, not in configuration files. See Connecting Models.
- Licensing reports usage to the licensing service over HTTPS. See Licensing.
- Backup, updates, and logs follow the same approach, described in each model's guide.