OCR Engines
DocIntel supports four OCR engines for processing scanned documents and images. Configure via the OCR__ENGINE environment variable.
Supported engines
| Engine | OCR__ENGINE value | Self-hosted | Notes |
|---|---|---|---|
| Tesseract | tesseract | ✅ Yes | Default; local, no API key needed |
| Azure Document Intelligence | azure_doc_intel | ❌ Cloud | Best for structured forms + tables |
| Google Cloud Vision | google_vision | ❌ Cloud | Strong multilingual support |
| AWS Textract | textract | ❌ Cloud | Best for US healthcare + financial forms |
Tesseract (default)
bash
OCR__ENGINE=tesseract
OCR__LANGUAGES=en- Runs locally with no API key required
- Supports 100+ languages via traineddata files
- Ideal for air-gapped and development environments
- For multilingual documents, set
OCR__LANGUAGES=en,de,fr
Azure Document Intelligence
bash
OCR__ENGINE=azure_doc_intel
OCR__AZURE_ENDPOINT=https://your-resource.cognitiveservices.azure.com/
OCR__AZURE_KEY=your-key- Best accuracy for structured forms, tables, and checkboxes
- Automatic table structure recognition
- Pre-built models for invoices, receipts, and ID documents
Google Cloud Vision
bash
OCR__ENGINE=google_vision
OCR__GOOGLE_CREDENTIALS_JSON=/path/to/credentials.json- Strongest multilingual text detection
- Handles mixed-script documents well
- Good handwriting recognition
AWS Textract
bash
OCR__ENGINE=textract
OCR__AWS_REGION=us-east-1- Uses IAM credentials or instance profile for authentication
- Specialized for US healthcare (HIPAA) and financial forms
- Strong table and key-value extraction
Vision engines (sketches & whiteboard)
For hand-drawn sketches, architectural diagrams, and whiteboard photos, DocIntel uses a separate vision engine (not OCR):
| Engine | SKETCH__ENGINE value | Description |
|---|---|---|
| GPT-5 Vision | openai-vision | Default; best quality semantic understanding |
| Azure AI Vision 4.0 | azure-vision | Dense captioning + OCR |
| Google Cloud Vision | google-vision | Multi-feature analysis |
| Tesseract (local) | local | OCR only — offline fallback |
bash
SKETCH__ENGINE=openai-vision
SKETCH__DETAIL_LEVEL=highSee Configuration → for the full reference.
License tier requirements
| Feature | Beta | Pro | Enterprise |
|---|---|---|---|
| Tesseract OCR | ✅ | ✅ | ✅ |
| Azure Document Intelligence | ❌ | ✅ | ✅ |
| Google Cloud Vision | ❌ | ✅ | ✅ |
| AWS Textract | ❌ | ✅ | ✅ |
Beta-tier licenses are limited to Tesseract OCR. See pricing →