Skip to main content
โ† Back to Gallery

๐Ÿ“„ Smart OCR

Resumable OCR uses the v3 text-quality correction pipeline before consumer-supplied structured JSON extraction.

Workflow Details

Workflow ID: smart_ocr_v1
Version: 1.5.1
Total Steps: 3

Workflow Steps

1
Image Extraction
python
โ†“
2
Ocr Extraction
python
โ†“
3
Structure Extraction
python

Input Configuration

Select an instance to apply configuration overrides.
Supported formats: PDF, DOC, DOCX, TXT, PNG, JPG
Apply safe OCR text corrections before JSON extraction
Repair likely encoding/mojibake artifacts
Digit correction mode: auto, on, or off
OCR model override for the ocr_extraction step; blank keeps the node's configured default (olmOCR-2-7B-1025).
Optional: rasterize PDF pages to 06_images/ up front, e.g. for preview/QA independent of OCR. ocr_extraction reuses whatever pages already exist there either way, so enabling this never causes double rendering.
Consumer-supplied Gemma JSON-schema prompt or manual JEV questions; leave empty when schema_name generates JEV questions.
Optional packaged schema for schema-aware JEV decisions; use this instead of a manual JEV question prompt.
Structured-extraction model. JEV returns TypeSafe answers and can derive questions from schema_name; it is not a general free-form JSON generator. Independent of ocr_model.

Execution Preview

This workflow will execute the following steps in order:

1 Image Extraction
Processor: python
โ†“
2 Ocr Extraction
Processor: python
โ†“
3 Structure Extraction
Processor: python
Total Steps: 3

Example Output

{
  "structured_data": {
    "sale_id": "RG 123/2024",
    "tribunal": "Tribunale di Roma",
    "lot_number": "1",
    "assets": [...],
    "valuation": {
      "market_value": 250000.00,
      "auction_base_price": 187500.00
    }
  },
  "metadata": {
    "processing_time": 24.7,
    "llm_provider": "mistralai/Mistral-7B-Instruct-v0.3"
  }
}