Skip to main content

​
Overview

The Extract endpoint combines document parsing with schema-based data extraction. Define a JSON schema describing the data you need, and Aifano extracts it from any document. Credits: 2 credits per page

​
Basic Usage

curl -X POST "https://platform.aifano.com/extract" \
  -H "Authorization: Bearer $AIFANO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input": "aifano://invoice.pdf",
    "schema": {
      "type": "object",
      "properties": {
        "invoice_number": { "type": "string" },
        "date": { "type": "string" },
        "total_amount": { "type": "number" },
        "line_items": {
          "type": "array",
          "items": {
            "type": "object",
            "properties": {
              "description": { "type": "string" },
              "quantity": { "type": "number" },
              "unit_price": { "type": "number" }
            }
          }
        }
      }
    }
  }'

​
Response

{
  "job_id": "job_xyz789",
  "duration": 4.56,
  "usage": { "num_pages": 2, "credits": 4 },
  "result": {
    "invoice_number": "INV-2025-001",
    "date": "2025-01-15",
    "total_amount": 1250.00,
    "line_items": [
      { "description": "Consulting Services", "quantity": 10, "unit_price": 100.00 },
      { "description": "Software License", "quantity": 1, "unit_price": 250.00 }
    ]
  }
}

​
Schema Design Tips

Property names guide the extraction model. invoice_number works better than id.
{
  "total_amount": {
    "type": "number",
    "description": "The total invoice amount including tax"
  }
}
Line items, table rows, and lists should use array type with items schema.
{
  "status": {
    "type": "string",
    "enum": ["paid", "pending", "overdue"]
  }
}

​
Custom System Prompt

Guide the extraction model with a custom system prompt: 
{
  "input": "aifano://contract.pdf",
  "schema": { "..." : "..." },
  "system_prompt": "Extract all monetary values in EUR. Dates should be in ISO 8601 format."
}

​
Reusing Parsed Results

If you’ve already parsed a document, pass the job_id to skip re-parsing: 
{
  "input": "jobid://job_abc123",
  "schema": { "..." : "..." }
}
This saves credits by reusing the parse result (you only pay extraction credits, not parse credits).

​
Parsing Options

Control how the document is parsed before extraction: 
{
  "input": "aifano://document.pdf",
  "schema": { "..." : "..." },
  "parsing": {
    "enhance": { "agentic": [{ "scope": "table" }] },
    "settings": { "ocr_system": "standard" }
  }
}
See Parse for all available parsing options.