stax-ai-module-sdk 0.1.9

Stax.ai Module SDK

Stax.ai Module SDK

This project is created and maintained by Stax.ai, Inc., and is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License.

About

...coming soon...

Installation

pip install stax_module_sdk

Usage

Create a Stax.ai module

With appropriate developer API credentials, POST the following data to https://api.stax.ai/module/create to create a module:

{
    "name": "Module name",
    "description": "Explain what the module does...",
    "rate": 0.05,
    "config": []
}

The rate property is the number of US dollars charged per unit recorded by the module. The unit can be a single document, number of pages processed, number of data points extracted, etc.

The config property is a list of module configuration objects. A config element has the format:

{
    "label": "User-readable configuration label",
    "dataType": "string",
    "required": true,
    "default": "Hello world",
    "hint": "Tell the user what this field means",
    "options": [],
}

The dataType can be one of :

• string (plain text field),
• list to have the user pick an option from a provided list,
• stack to have the user pick a stack from their team, and
• file to let the user upload a configuration file.

The required property defines whether a user needs to input this configuration parameter.

The default value is prepopulated in the field.

The hint is a helper that displayes below the field assisting the user in getting the information needed.

The options property is a list of strings used as dropdown menu options if the dataType is list.

Upon creating a module, you will receive the module ID and a key, which you must use to authenticate any calls from the module function.

You can also update the module by POSTing any changes to https://api.stax.ai/module/update:

{
    "moduleId": "UNIQUE_MODULE_ID",
    "moduleDiff": {
        "name": "New module name"
    }
}

Write your module application

import os
from stax_module_sdk import def_module, Stax

# Load module ID and key (which you got when creating the module) from environment variables or hardcode those here
MODULE_ID = os.getenv('MODULE_ID')
MODULE_KEY = os.getenv('MODULE_KEY')

@def_module(MODULE_ID, MODULE_KEY)
def main(stax:Stax, document:str, stack:str, team:str, config:list[dict]):
    '''
    Module document processing function. Is provided the following arguments:
    - stax [stax_module_sdk.Stax]: Stax.ai API object instantiated with team and module auth
    - document [str]: Document ID string
    - stack [str]: Stack ID string
    - team [str]: Team ID string
    - config [List(stax_module_sdk.Config)]: Pipeline configuration for module/step

    Return the number of units to charge for (pages? items extracted? documents?)
    '''

    # Load document
    doc = stax.getDocument()

    # Put your module functionality here
    # ...

    # Update document with difference/changes
    stax.updateDocument(diff={})

    # Raise an exception to stop the pipeline and flag the document
    raise Exception("Oops, something went wrong!")

    return 1


# Call your module function as the main function
if __name__ == "__main__":
    main()
    # You don't need to pass any arguments in. The @def_module decorator
    #  takes care of loading everything from your module credentials and
    #  the request.json file used to load the module

    # You can call the main function with the test kwarg set to True
    #  during development to prevent Stax.ai API call attempts.
    main(test=True)

Deploy your module

Publish your module

"Stax" Class Reference

Constructor

stax = Stax(MODULE_ID, MODULE_KEY, TEAM_ID, document=DOCUMENT_ID)

Get Document

doc = stax.getDocument(document=DOCUMENT_ID)

If you don't pass in the document keyword argument and are calling this within a module definition, the document that the module is currently processing will be loaded.

Download Document PDF

stax.downloadDocument(document=DOCUMENT_ID, path='/path/to/file.pdf')

Document Document Page

stax.downloadPage(document=DOCUMENT_ID, path='/path/to/page.pdf', key=PAGE_KEY)

Update Document

doc = stax.updateDocument(document=DOCUMENT_ID, diff={ ... })

Download configured file

stax.downloadConfigFile(config, CONFIG_LABEL)

GET from Stax.ai API

You can also make any Stax.ai API call using the module-authenticated Stax object. See the Stax.ai Developer API reference for more information.

res = stax.get('/document/get?docId=' + DOCUMENT_ID)

POST to Stax.ai API

res = stax.post('/document/update', {
    "docId": DOCUMENT_ID,
    "docDiff": {}
})

Helper functions

The submodule stax_module_sdk.helpers includes a few helper functions which are used quite often in Stax.ai modules.

Get config value by label

from stax_module_sdk.helpers import getConfigByLabel

value = getConfigByLabel(config, label)

Get document field by key

from stax_module_sdk.helpers import getFieldByKey

field = getFieldbyKey(doc["metadata"], key)

Get document fields (multiple) by key

from stax_module_sdk.helpers import getFieldsByKey

fields = getFieldsbyKey(doc["metadata"], key)

Find document field index by key and value

from stax_module_sdk.helpers import getFieldIndex

fieldIdx = getFieldIndex(doc["metadata"], key, value)

PyText Parser

Stax.ai accepts the PyText format (it's custom) in various modules where the config dataType is set to "pytext". In this format, users can run custom python commands within { ... } in their text. This can be used to provide easy configuration of messages, etc. Anything within the { ... } characters are evaluated with the following properties available:

• the 'doc' variable containers the document resource,
• the 'field' function which gets the value of a field provided a key,
• the 'link' variable is a link to the document.

In order to use the built-in PyText parser, follow the example below:

from stax_module_sdk.helpers import getConfigByLabel
from stax_module_sdk.pytext import renderPyText

pt = getConfigByLabel(config, "Message Body")
text = renderPyText(doc, pt)

For SDK Developers

Once updates are made, increment the version number in setup.py and push to pypi.

python setup.py sdist bdist_wheel
twine upload --repository-url https://test.pypi.org/legacy/ dist/*
twine upload dist/*

NOTE: PyPi username is: __token__