This repository contains the official Hydra API Python client. The Hydra API is a text recognition service.

1. Install the package.

pip install hydra-api

or

poetry add hydra-api

etc.

1. Create a new data source on siftrics.com.
2. Grab an API key from the page of your newly created data source.
3. Create a client, passing your API key into the constructor.
4. Use the client to processes documents, passing in the id of a data source and the filepaths of the documents.

import hydra_api

client = hydra_api.Client('xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx')

rows = client.recognize('my_data_source_id', ['invoice.pdf', 'receipt_1.png'])

rows looks like this:

[
  {
    "Error": "",
    "FileIndex": 0,
    "RecognizedText": { ... }
  },
  ...
]

FileIndex is the index of this file in the original request's "files" array.

RecognizedText is a dictionary mapping labels to values. Labels are the titles of the bounding boxes drawn during the creation of the data source. Values are the recognized text inside those bounding boxes.

There is another function, client.recognizeBase64(dataSourceId, base64Files, doFaster=False), which accepts base64 strings (file contents) instead of file paths. Because it is not trivial to infer MIME type from the contents of a file, you must specify the MIME type associated to each base64 file string: base64Files must be a list of dict objects containing two fields: "mimeType" and ``"base64File"`. Example:

    base64Files = [
        {
            'mimeType': 'image/png',
            'base64File': '...',
        },
        {
            'mimeType': 'application/pdf',
            'base64File': '...',
        },
    ]
    rows = client.recognizeBase64('Helm-Test-Againe', base64Files, doFaster=True)

Hydra can transform input documents so they are cropped and aligned with the original image used to create the data source.

The recognize and recognizeBase64 functions have an additional default parameter, returnTransformedImages, which defaults to False, but if it's set to True then Siftrics transforms and returns images so they are aligned with the original image.

Returned images will be available in the "TransformedImages" field of each element of "Rows" in the response:

{
  "Rows": [
    {
      "Error": "",
      "FileIndex": 0,
      "RecognizedText": {
        "My Field 1": "text from your document...",
        "My Field 2": "text from your document...",
        ...
      },
      "TransformedImages": [
        {
          "Base64Image": ...,
          "PageNumber": 1
        },
        ...
      ]
    },
    ...
  ]
}

The recognize and recognizeBase64 functions have an additional default parameter, doFaster, which defaults to False, but if it's set to True then Siftrics processes the documents faster at the risk of lower text recognition accuracy. Experimentally, doFaster=true seems not to affect accuracy when all the documents to be processed have been rotated no more than 45 degrees.

The recognize and recognizeBase64 functions have additional default parameters, returnJpgs=False and jpgQuality=85. If returnJpgs is set to True, then Siftrics returns cropped images in JPG format instead of PNG format. jpgQuality must be an integer between 1 and 100 inclusive.

Here is the official documentation for the Hydra API.

This code is licensed under Apache V2.0. The full text of the license can be found in the "LICENSE" file.