In this example, we’ll learn how to apply optical character recognition (OCR) to an invoice document.

​
Prerequisites

You will need:

  • An ABBYY developer account
  • An API key
  • If using Python SDK: Python v3.8+
  • If using TypeScript SDK: ES2018+, Node.js active and maintenance LTS releases (currently, v18 and v20), Bun v1+, Deno v1.39+
  • An invoice document to process (or use one of ours!)

Create a developer account to get access to the API. It’s free to get started. After signing up, navigate to the API Keys page and create an API key.

​
Document to process

If you have an invoice handy, feel free to try it out with this Quickstart. Otherwise, we’ll use this Canadian invoice hosted on GitHub.

​
Install the SDK

While you can call the API directly, the Document AI API provides SDKs for several popular programming languages. If you don’t have a project already, create one:

# Create a virtual environment
python3 -m venv venv
# On macOS or Linux:
source venv/bin/activate
# On Windows:
venv\Scripts\activate
# Create a file
touch docai.py

Next, install the Document AI API SDK:

pip install abbyy-document-ai

​
Call the API

Call the API using the documents/invoice endpoint. Provide the path to the invoice file to process (or use the one we’ve provided) and insert your API key into the Authorization header.

from document_ai_sdk import DocumentAi

API_KEY="your-api-key-here"
with DocumentAi(api_key_auth=API_KEY) as document_ai:
res = document_ai.invoice.extract_invoice_data(input_source={
"url": "https://raw.githubusercontent.com/dotNetkow/test-cods/6c25fa2e54fb96ecf534903aef0d3cbecd71e987/Invoice%20CA_1.pdf",
})

docId = res.documents[0].id
print(res)

Run the code using python3 docai.py or node docai.ts. The API will respond with a document id that we’ll need in the next step.

[
  {
    "id": "r23qpmmjsc7f93o3",
    "name": "Invoice%20CA_1.pdf",
    "createdAt": "2025-04-01 20:41:26.510152",
    "status": "Pending",
    "pageCount": 1
  }
]

​
Retrieve the Results

The Document AI API begins processing the invoice immediately on submission. To retrieve the results, we make one more API call to /documents/invoice/[document-id]. Fill in the document-id using the id from the previous API call. The API’s processing time varies based on a variety of factors including number of document pages, so we’ll keep attempting to retrieve the extracted fields every few seconds until the API’s status returns Processed.

# import time module at the top of the file for waiting between field checks
import time

processed = False
while not processed:
time.sleep(3) # Wait 3 seconds between checks
res = document_ai.models.invoice.get_extracted_fields(
document_id=doc_id
)
processed = res.invoice.meta.status == "Processed"

print("Invoice number: ", res.invoice.fields.invoice_number)

Success! We now have all the fields from the invoice document:

Invoice fields
{
  "meta": {
    "id": "r23qpmmjsc7f93o3",
    "name": "Invoice%20CA_1.pdf",
    "createdAt": "2025-04-01 20:41:26.510152",
    "status": "Processed",
    "pageCount": 1
  },
  "fields": {
    "invoiceNumber": "9435435",
    "invoiceDate": "2021-11-11",
    "total": 4620,
    "currency": "CAD",
    "businessUnit": {
      "name": "M GROUP INC.",
      "address": "770-21ST AVENUE N\nSASKATOON SK S1K7R1\nCANADA",
      "country": "CA"
    },
    "vendor": {
      "name": "ANADAYA CANADA",
      "address": "262 Merrier Avenue, Toronto, Ontario Т1Т 3R29"
    },
    "reversedCharge": false,
    "invoiceType": {
      "invoice": true,
      "creditNote": false
    },
    "purchaseOrder": [
      {
        "orderChecked": false
      }
    ],
    "lineItems": [
      {
        "isValid": false,
        "position": 1,
        "articleNumberVendor": "864UW7",
        "description": "RECEIVER (19003)",
        "quantity": 2,
        "unitPrice": 385,
        "netPrice": 770,
        "currency": "CAD"
      },
      {
        "isValid": false,
        "position": 2,
        "articleNumberVendor": "864UW7",
        "description": "RECEIVER (85813)",
        "quantity": 2,
        "unitPrice": 385,
        "netPrice": 770,
        "currency": "CAD"
      },
      {
        "isValid": false,
        "position": 3,
        "articleNumberVendor": "384 UW5",
        "description": "RECEIVER (19053)",
        "quantity": 2,
        "unitPrice": 385,
        "netPrice": 770,
        "currency": "CAD"
      },
      {
        "isValid": false,
        "position": 4,
        "articleNumberVendor": "365UW8",
        "description": "RECEIVER (85583)",
        "quantity": 2,
        "unitPrice": 385,
        "netPrice": 770,
        "currency": "CAD"
      },
      {
        "isValid": false,
        "position": 5,
        "articleNumberVendor": "354UW7",
        "description": "RECEIVER (15003)",
        "quantity": 2,
        "unitPrice": 385,
        "netPrice": 770,
        "currency": "CAD"
      },
      {
        "isValid": false,
        "position": 6,
        "articleNumberVendor": "564UW7",
        "description": "RECEIVER (85513)",
        "quantity": 2,
        "unitPrice": 385,
        "netPrice": 770,
        "currency": "CAD"
      }
    ]
  }
}

​
Complete Code

Here’s the complete code for reference.

import time
from document_ai_sdk import DocumentAi

API_KEY="your-api-key-here"
with DocumentAi(api_key_auth=API_KEY) as document_ai:
res = document_ai.invoice.extract_invoice_data(input_source={
"url": "https://raw.githubusercontent.com/dotNetkow/test-cods/6c25fa2e54fb96ecf534903aef0d3cbecd71e987/Invoice%20CA_1.pdf",
})

docId = res.documents[0].id
print(res)

processed = False
while not processed:
time.sleep(3) # Wait 3 seconds between checks
res = document_ai.models.invoice.get_extracted_fields(
document_id=doc_id
)
processed = res.invoice.meta.status == "Processed"

print("Invoice number: ", res.invoice.fields.invoice_number)

​
Next Steps

Now that you’ve made your first Document AI API request, it’s time to explore what else is possible.

Explore the Models

Review all document models and their capabilities for extracting data

Try our API

Learn how to use the Document AI API to process documents programmatically

Join the Community

Exchange ideas, get expert guidance, and enhance your OCR, IDP, and AI implementations

Was this page helpful?