GitHub Content Moderator Action

Overview

This GitHub Action moderates the content of issues, pull requests, and comments using the OpenAI Content Moderation API. It also supports OpenAI-compatible custom endpoints.

Features

• Automatic Text Extraction: Automatically includes titles for Issues, PRs, and Discussions
• Text Normalization: Normalizes text for consistent detection (Unicode NFKC, whitespace handling)
• Smart Content Hiding:
  • Comments: Minimized (collapsed with reason)
  • Issues: Closed + Locked
  • Pull Requests: Closed + Locked
  • Discussions: Locked
• Custom Endpoint Support: Works with GitHub Copilot API, OpenAI, Azure OpenAI, Ollama, and any OpenAI-compatible endpoint

Usage

Here is an example of how to use this action in your workflow:

name: Moderate Content

on:
  issues:
    types: [opened, edited]
  pull_request:
    types: [opened, edited]
  issue_comment:
    types: [created, edited]
  pull_request_review_comment:
    types: [created, edited]
  discussion:
    types: [created, edited]
  discussion_comment:
    types: [created, edited]

jobs:
  moderate:
    runs-on: ubuntu-latest
    permissions:
      issues: write
      pull-requests: write
      discussions: write
    steps:
      - name: Moderate content
        id: moderator
        uses: Walkmana-25/github-comment-moderation@main
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          openai-api-key: ${{ secrets.OPENAI_API_KEY }}
          openai-endpoint: ${{ secrets.OPENAI_ENDPOINT }}
          openai-model: ${{ secrets.OPENAI_MODEL }}
          # text-to-moderate is optional - auto-populated from event

      - name: Post-moderation summary
        if: steps.moderator.outputs.is-inappropriate == 'true'
        run: |
          echo "Content was flagged for the following reasons: ${{ steps.moderator.outputs.flagged-categories }}"

Configuration Options

This action supports the following inputs:

Input	Required	Default	Description
github-token	Yes	${{ github.token }}	The GitHub token for API authentication
openai-api-key	No	${{ github.token }}	The API key for the OpenAI-compatible endpoint. If not provided, the github-token will be used (for GitHub Copilot API)
openai-endpoint	No	https://models.github.ai/inference	The base URL of the OpenAI-compatible API endpoint
openai-model	No	openai/gpt-4.1-mini	The model to use for content moderation
text-to-moderate	No	(auto-populated)	The text content to moderate. If not provided, will be auto-populated from the event:
			- Issues/PRs/Discussions: title + "\\n\\n" + body
			- Comments: body only
retry-count	No	3	Number of retry attempts for API calls

Content Moderated by Event Type

Event Type	Content Moderated	Action When Flagged
issues	Title + Body	Closed + Locked
pull_request, pull_request_target	Title + Body	Closed + Locked
issue_comment	Comment body	Minimized
pull_request_review_comment	Comment body	Minimized
discussion	Title + Body	Locked
discussion_comment	Comment body	Minimized

Custom Endpoint Usage

This action supports OpenAI-compatible endpoints, allowing you to use various AI providers:

GitHub Copilot API (Default)

with:
  github-token: ${{ secrets.GITHUB_TOKEN }}
  # openai-api-key defaults to github-token
  # openai-endpoint defaults to https://models.github.ai/inference
  openai-model: 'openai/gpt-4.1-mini'

OpenAI API

with:
  github-token: ${{ secrets.GITHUB_TOKEN }}
  openai-api-key: ${{ secrets.OPENAI_API_KEY }}
  openai-endpoint: 'https://api.openai.com/v1'
  openai-model: 'gpt-4o-mini'

Azure OpenAI Service

with:
  github-token: ${{ secrets.GITHUB_TOKEN }}
  openai-api-key: ${{ secrets.AZURE_OPENAI_API_KEY }}
  openai-endpoint: 'https://your-resource.openai.azure.com/'
  openai-model: 'gpt-4'  # Use the deployment name, not the model name

Ollama (Local LLM)

with:
  github-token: ${{ secrets.GITHUB_TOKEN }}
  openai-api-key: 'ollama'  # Ollama may not require an API key
  openai-endpoint: 'https://your-ollama-instance.com/v1'
  openai-model: 'llama3.2'

Any OpenAI-Compatible API

with:
  github-token: ${{ secrets.GITHUB_TOKEN }}
  openai-api-key: ${{ secrets.CUSTOM_API_KEY }}
  openai-endpoint: 'https://your-custom-endpoint.com/v1'
  openai-model: 'custom-model-name'

Outputs

Output	Description
is-inappropriate	Whether the content was flagged as inappropriate (true or false)
flagged-categories	Comma-separated list of flagged categories (e.g., hate,violence)
moderation-results-json	Full JSON response from the API

Text Normalization

The action automatically normalizes text before moderation to ensure consistent detection:

• Unicode NFKC normalization: Handles compatibility characters (important for Japanese text)
• Whitespace normalization: Collapses multiple spaces/tabs into single spaces
• Trailing newline removal: Ensures "text\n" is detected the same as "text"

This means variations like "ばーかばーか\n" will be detected consistently as inappropriate.

Permissions

The action requires the following permissions:

permissions:
  issues: write      # For closing/locking issues
  pull-requests: write  # For closing/locking PRs
  discussions: write   # For locking discussions

Content Policy

The action uses the following content policy for moderation:

• Hate speech or discriminatory language
• Sexual content or explicit material
• Violence, threats, or promoting harm
• Self-harm or suicide-related content
• Harassment or bullying