> ## Documentation Index
> Fetch the complete documentation index at: https://docs.helloatlas.in/llms.txt
> Use this file to discover all available pages before exploring further.

# IDP Extraction

> Extract structured data from unstructured documents using Atlas IDP APIs. Upload documents, retrieve extracted fields, and receive results through polling or callbacks.

Atlas IDP Extraction APIs allow you to submit documents for OCR processing and receive structured JSON outputs containing extracted fields, confidence scores, document classifications, and validation metadata.

<Note>
  **Base URL**

  `https://docstream.dev.kreditmind.com`

  All endpoints in this guide are relative to this base URL.
</Note>

<Steps>
  <Step title="Generate an access token">
    Authenticate using your assigned credentials.

    ### Endpoint

    ```http theme={null}
    POST /v1/docstream/authtoken
    ```

    <CodeGroup>
      ```bash cURL theme={null}
      curl --request POST \
        --url https://docstream.dev.kreditmind.com/v1/docstream/authtoken \
        --header "Client-Id: your-client-id" \
        --header "Client-Secret: your-client-secret"
      ```

      ```json Response theme={null}
      {
        "access_token": "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...",
        "expires_in": 3600
      }
      ```
    </CodeGroup>

    The returned `access_token` must be passed as the `Token` header in all subsequent requests.
  </Step>

  <Step title="Upload documents for extraction">
    Submit one or more document URLs for processing.

    ### Endpoint

    ```http theme={null}
    POST /v1/docstream/multiupload
    ```

    ### Request fields

    | Field                     | Type   | Required | Description                      |
    | ------------------------- | ------ | -------- | -------------------------------- |
    | `product_type`            | string | Yes      | Product type assigned by Atlas   |
    | `file_urls`               | array  | Yes      | List of documents to process     |
    | `file_urls[].file_url`    | string | Yes      | Pre-signed URL of the document   |
    | `file_urls[].document_id` | string | Yes      | Your internal document reference |

    <CodeGroup>
      ```bash cURL theme={null}
      curl --request POST \
        --url https://docstream.dev.kreditmind.com/v1/docstream/multiupload \
        --header "Token: your-access-token" \
        --header "Content-Type: application/json" \
        --data '{
          "product_type": "LOAN_APP",
          "file_urls": [
            {
              "file_url": "https://your-bucket.s3.ap-south-1.amazonaws.com/pan.jpg",
              "document_id": "001"
            },
            {
              "file_url": "https://your-bucket.s3.ap-south-1.amazonaws.com/aadhaar.jpg",
              "document_id": "002"
            }
          ]
        }'
      ```

      ```json Response theme={null}
      {
        "batch_id": "BATCH-30732-SIYKJ-20251121113228"
      }
      ```
    </CodeGroup>

    Atlas immediately starts processing the uploaded documents.
  </Step>

  <Step title="Poll extraction results">
    Retrieve extracted OCR data using the returned `batch_id`.

    ### Endpoint

    ```http theme={null}
    GET /v1/docstream/extracts
    ```

    ### Query Parameters

    | Parameter  | Required | Description                           |
    | ---------- | -------- | ------------------------------------- |
    | `batch_id` | Yes      | Batch identifier returned from upload |
    | `page`     | No       | Page number (default: 1)              |
    | `per_page` | No       | Number of results per page            |

    <CodeGroup>
      ```bash cURL theme={null}
      curl --request GET \
        --url "https://docstream.dev.kreditmind.com/v1/docstream/extracts?batch_id=BATCH-30732-SIYKJ-20251121113228&page=1&per_page=50" \
        --header "Token: your-access-token"
      ```

      ```json Response theme={null}
      {
        "data": [
          {
            "document_id": "001",
            "file_data": {
              "document_type": "PAN",
              "name_as_per_pan": {
                "value": "John Doe",
                "confidence_score": 0.99
              },
              "pan_number": {
                "value": "ABCDE1234F",
                "confidence_score": 0.98
              },
              "dob": {
                "value": "17/10/1991",
                "confidence_score": 0.97
              }
            }
          }
        ],
        "page": 1,
        "per_page": 50,
        "total_pages": 1,
        "total_items": 1,
        "has_next": false,
        "has_prev": false
      }
      ```
    </CodeGroup>
  </Step>
</Steps>

## Callback Support

Atlas supports asynchronous result delivery through callbacks.

Share the following details with the Atlas team during onboarding:

* `callback_url` — Endpoint where extracted results should be delivered
* `auth_url` — Endpoint Atlas can call to obtain authentication credentials before posting results

Atlas maintains the callback configuration on our end, so these URLs only need to be shared once during setup.

Once processing is complete, Atlas will automatically POST the extraction results to your configured callback endpoint.

<Note>
  Callback delivery is recommended for production environments as it eliminates the need for continuous polling and provides results as soon as processing is completed.
</Note>

### Example callback payload

```json theme={null}
{
  "data": [
    {
      "document_id": "001",
      "file_data": {
        "document_type": "AADHAAR",
        "name": {
          "value": "Yarlagadda Vijaya Lakshmi",
          "confidence_score": 1
        },
        "date_of_birth": {
          "value": "16/04/1983",
          "confidence_score": 1
        },
        "address": {
          "value": "W/O: Yarlagadda Siva Rama Krishna...",
          "confidence_score": 1
        }
      }
    }
  ],
  "page": 1,
  "per_page": 50,
  "total_pages": 1,
  "total_items": 5,
  "has_next": false,
  "has_prev": false
}
```

Refer to the **Supported Documents** section for the complete field-level extraction reference for every document type.