Order Table Extraction

In this example, we are a company that sells many different products. Our customer sends us a small table of all the items they want to buy. However, since the customer creates the table, it can look slightly different every time. For example, they write Product ID instead of product_id, Qty instead of Quantity, and the shipping times have no clear structure.

Product ID	Qty	Unit price	Manufacturar	Shipping Time
TZX22-EHZ2	100	2.76$	Tusp	1-2 days
ZUI23-772L6	250	5.00$	-	1 week
UIUU-13BMW	340'000	0.001$	puma	About a month
QUE2-AIME2	56	45.56$	lebra	Tomorrow

7KB

OrderTable.docx

Waveline Extract makes it easy to unify these fields into one format we define.

Let's construct a Shape to extract product_id, unit_price, quantity, and shipping_time for each product:

[
  {
    "name": "products",
    "type": "object",
    "description": "All products from the table",
    "isArray": true,
    "elements": [
      {
        "name": "product_id",
        "type": "string",
        "description": "The id of that product. aka product number",
        "isArray": false
      },
      {
        "name": "quantity",
        "type": "number",
        "description": "Quantity of how many units. Aka Qty",
        "isArray": false
      },
      {
        "name": "unit_price",
        "type": "number",
        "description": "Unit price of that product in dollars",
        "isArray": false
      },
      {
        "name": "shipping_time",
        "type": "string",
        "description": "Time it takes to ship this product. (In days)",
        "isArray": false
      }
    ]
  }
]

We can now call the /extract-document endpoint with this shape and the table as the payload to create the job:

curl -X POST "https://waveline.ai/api/v1/extract-document" \
     -H "Content-Type: application/json" \
     -H "Authorization: Bearer YOUR_API_KEY" \
     -d '{
          "fileName": "OrderTable.txt",
          "contentType": "application/pdf",
          "base64Content": "JVBERi0xLjMKMSAwIG9iago8PC9UeXBlL0NhdGF...",
          "shape": YOUR_SHAPE
        }'

After some time, we query for the result of this job with the job endpoint and get the following in the result field:

{
  "products": [
    {
      "product_id": "TZX22-EHZ2",
      "quantity": "100",
      "unit_price": 2.76,
      "shipping_time": "1-2 days"
    },
    {
      "product_id": "ZUI23-772L6",
      "quantity": "250",
      "unit_price": 5,
      "shipping_time": "1 week"
    },
    {
      "product_id": "UIUU-13BMW",
      "quantity": "340000",
      "unit_price": 0.001,
      "shipping_time": "About a month"
    },
    {
      "product_id": "QUE2-AIME2",
      "quantity": "56",
      "unit_price": 45.56,
      "shipping_time": "Tomorrow"
    }
  ]
}

As we can see above, all the fields have successfully been unified into the format we defined!