biohazard-alert

A GitHub App built with Probot to deliver notifications of toxic comments

Description

Listens for new or edited issues, pull requests, or comments. It sends the content of those events to a semantic analysis API that can rate the content on multiple sentiment axes. If the content is rated above a threshold on any axis, a notification email is sent to humans to investigate and decide whether to take action.

Configuration

This Probot app reads its configuration from two files:

• Global settings: .github repository under the user or organization it is installed in from the .github/biohazard-alert.yml file
• Repo-specific settings: .github/biohazard-alert.yml file

Configuration settings are:

• notifyOnError: true means that notifications are generated when errors are encountered (default true)
• skipPrivateRepos: true means that events from private repositories will be ignored (default true)
• threshold: Analysis ratings higher than this number will generate notifications (default 0.8)

Analysis

This app uses Google's Perspective API to analyze the content using the following models:

• TOXICITY
• SEVERE_TOXICITY
• IDENTITY_ATTACK
• INSULT
• PROFANITY
• THREAT
• SEXUALLY_EXPLICIT
• FLIRTATION
• UNSUBSTANTIAL

Development

# Install dependencies
npm install

# Build the app
npm run build

# Run the bot locally
npm run dev

License

MIT