Deployment Options
DocAI Fabric offers flexible deployment options to fit your infrastructure requirements, security policies, and operational preferences.
Choose Your Deployment Model
| Option | Best For | Infrastructure | Managed By |
|---|---|---|---|
| SaaS | Fastest start, no infra to manage | Our cloud | Us |
| Managed in Azure | New subscription inside your private network, we handle everything | Your Azure | Us (everything) |
| Self-Hosted (On-Premises) | Full control, on-premises (standalone or highly available) | Your servers | You |
| Managed in AWS | AWS-native environments | Your AWS | Us (deployment & updates) |
| GCP | GCP-native environments | Your GCP | You |
Comparison
| Feature | SaaS | Managed Azure | Managed AWS | Self-Hosted | GCP |
|---|---|---|---|---|---|
| Setup time | Minutes | Days (your network team prepares subnets and DNS) | ~1 hour | ~2 hours | ~2 hours |
| Data location | Our Azure tenant | Your Azure subscription | Your AWS account | Your infrastructure | Your cloud |
| Updates | Automatic | Managed by us | Managed by us | Manual | Manual |
| Custom domain | Included | Supported | Supported | Your responsibility | Your responsibility |
| Scaling | Automatic | Automatic (Container Apps) | Automatic (ECS) | Manual | Manual |
| AI services | Included | Your Azure OpenAI | AWS Bedrock + Textract (your account) | Your endpoints | Your endpoints |
| Monitoring | Included | Your Application Insights | Your CloudWatch | Your tools | Your tools |
What Every Deployment Needs
Regardless of deployment model, DocAI Fabric requires:
- An LLM for classification and extraction: Azure OpenAI (GPT-4.1-mini or newer recommended), Anthropic Claude on Amazon Bedrock for AWS deployments (in development), or any OpenAI-compatible endpoint you connect on the Models page (including your own on-premises GPU servers for self-hosted deployments)
- An OCR service: Azure Document Intelligence (Read model), Amazon Textract for AWS deployments (in development), or the built-in PaddleOCR and Tesseract engines that ship with self-hosted deployments and need no cloud service
- Container runtime: Docker or a managed container service
- Redis: for job queue and caching
- Blob / file storage: for document persistence