DocumentCloud GPT-3 Add-On

Warning: This Add-On is tied to a paid service and can cost money. Please use on batches of 10-20 documents without prior discussion. It analyzes the first page of documents and categorizes/summarizes them across a variety of criteria.

Files

addon.py

This file is part of the python-documentcloud library.

This file contains a base class AddOn, which implements shared functionality for all DocumentCloud Add-Ons to use. In most cases, you should not need to edit this file. You will subclass this class in main.py.

Upon initializing this class, it parses the JSON passed in as an argument, and populates a number of member variables.

• client - A DocumentCloud client. This is a python library (https://github.com/MuckRock/python-documentcloud) allowing easy access to the DocumentCloud API. It will be configured with the access token passed in, which gives you access to the API as the user who activated the Add-On for 5 minutes. (NOTE: we may need a way to pass in a refresh token if Add-Ons need to run for more than 5 minutes)

• id - a UUID to identify this run. This is used to update progress, status and upload files for this particular run of the Add-On. This will be None if called from the test_addon.py script.

• documents - A list of document IDs selected when the Add-On was activated

• query - The search query which was active when the Add-On was activated

• user_id - The user ID of the user who activated this run of the Add-On.

• org_id - The organization ID of the active organization of the user who activated this run of the Add-On.

• data - The Add-On specific data.

There are also some methods which provide useful functionality for an Add-On.

• set_progress(self, progress) - This takes a single integer argument between 0 and 100 which represents the percent of progress the Add-On run has made, to inform the user of the progress. As it takes some time to be shown to the user, this is primarily of use for long running Add-Ons.

• set_message(self, message) - Takes a string of max length 255 characters. This sets a status message to let the user know what the status of the Add-On run is. Similar to set_progress, it is mostly useful for long running Add-Ons.

• upload_file(self, file) - Takes a file object to attach to this Add-On run. This will be presented to the user for download. This is useful for Add-Ons which want to return data such as a CSV file or other exports of data to the user. It is currently limited to one file per run, so please ZIP your files if you need to return more than one. The file will be available for download for five days, after which it will be permanently deleted from the server.

• send_mail(self, subject, content) - This is used to email yourself at the email address associated with your DocumentCloud account. This can be used to send a notification when an Add-On run is complete or just to send additional information to the user who ran the Add-On. It takes two character strings, one for the subject and one for the body content of the email. The content is plain text and does not currently support Markdown or HTML.

The script also accepts command line options to allow for easier testing for development purposes. It requires your DocumentCloud username and password if the add-on requires authentication, which is used to fetch a refresh and access token. They can be passed in as command line arguments (--username and --password), or in environment variables (DC_USERNAME and DC_PASSWORD).

You can also pass in a list of document IDs (--documents), a search query (--query), and JSON parameters for your Add-On (--data) - be sure to properly quote your JSON at the command line.

Example invocation:

python main.py --documents 123 --data '{"regex": [pattern1, pattern2]}'

python main.py --query "+user:john-doe-111111" --data '{"regex": [pattern1, pattern2]}'

main.py

This is the file to edit to implement your Add-On specific functionality. You should define a class which inherits from AddOn from addon.py. Then you can instantiate a new instance and call the main method, which is the entry point for your Add-On logic. You may access the data parsed by AddOn as well as using the helper methods defined there. The HelloWorld example Add-On demonstrates using many of these features.

If you need to add more files, remember to instantiate the main Add-On class from a file called main.py - that is what the GitHub action will call with the Add-On parameters upon being dispatched.

config.yaml

This is a YAML file which defines the data your Add-On expects to receive. DocumentCloud will use it to show a corresponding form with the proper fields. It uses the JSON Schema format, but allows you to use YAML for convenience. You may read more about JSON Schema, but here are the basics to get started:

# The title is the title of your Add-On
title: Hello World
# The description will be shown above the form when activating the Add-On
description: This is an updated simple test add-on
# Type should always be object
type: object
# Properties are the fields for your form
properties:
  # the key is the name of the variable that will be returned to your code
  name:
    # the title is what will be shown as the form label
    title: Name
    # a string is text
    type: string

At the top level you have the following properties:

• title - The title for your Add-On
• description - a description for your Add-On - will be displayed above the form when someone runs the add-on
• type - This should always be set to object
• properties - This is an object describing the data fields your add-on accepts
  • The name will be the name of the variable the data is returned in
  • title - The label shown on the form for this field
  • type - This may be string, number or boolean

requirements.txt

This is a standard pip requirements.txt file. It allows you to specify python packages to be installed before running the Add-On. You may add any dependencies your Add-On has here. By default we install the python-documentcloud API library and the requests HTTP request package. You may upgrade the python-documentcloud version when new releases come out in order to take advantage of new features.

.github/workflows/run-addon.yml

This is the GitHub Actions configuration file for running the add-on. It references a reusable workflow from the MuckRock/documentcloud-addon-workflows repository. This workflow sets up python, installs dependencies and runs the main.py to start the Add-On. It accepts two inputs:

• timeout - Number of minutes to time out. The default is 5. You may increase this if your add-on will run for longer than that.
• python-version - The version of python you would like to use. Defaults to 3.10.

To set an input:

jobs:
  Run-Add-On:
    uses: MuckRock/documentcloud-addon-workflows/.github/workflows/update-config.yml@v1
    with:
      timeout: 30

It is recommended you use the reusable workflow in order to receive future improvements to the workflow. If needed you may fork the reusable workflow and edit it as needed. If you do edit it, you should leave the first step in place, which uses the UUID as its name, to allow DocumentCloud to identify the run.

It would be possible to make a similar workflow for other programming languages if one wanted to write Add-Ons in a language besides Python.

.github/workflows/update-config.yml

This is the GitHub Actions configuration file for updating the configuration file. It references a reusable workflow from the MuckRock/documentcloud-addon-workflows repository. This workflow sends a POST request to DocumentCloud whenever a new config.yaml file is pushed to the repository. It accepts one input:

• url - The base URL for the DocumentCloud API. The default is "https://api.www.documentcloud.org/api/". It should only be changed if you are running your own instance of DocumentCloud.

LICENSE

The license this code is provided under, the 3-Clause BSD License

Reference

Full parameter reference

This is a reference of all of the data passed in to the Add-On. A single JSON object is passed in to main.py as a quoted string. The init function parses this out and converts it to useful python objects for your main function to use. The following are the top level keys in the object.

• token - An access token which will be valid for 5 minutes, giving you API access authorized as the user who activated the add-on. The init function uses this value to configure the DocumentCloud client object.

• refresh_token - A refresh token which will be valid for 1 day, giving you API access to new refresh tokens when they expire. The init function uses this value to configure the DocumentCloud client object.

• base_uri - This can be used to point the API server to other instances, such as our internal staging server. It should not be used unless you are running your own instance of DocumentCloud. It is also used in the initialization of the DocumentCloud client.

• auth_uri - The corresponding auth_uri if a base_uri is specified.

• documents - This is the list of Document IDs which is passed in to main

• query - This is the search query which is passed in to main

• data - This is the Add-On specific data, as defined when registering the Add-On with DocumentCloud. It is passed in to main in the params dictionary under the key data

• user and organization - The user ID and organiation ID of the user who activated the Add-On. They are also passed in to main through the params dictionary under the keys user and organization respectively.

• id - A UUID to uniquely identify this Add-On run. It allows DocumentCloud to identify the run, as well as allowing the run to send back progress, status message and file updates.