Repository files navigation

Github action that uses machine learning to detect potential toxic comments added to PRs and issues so authors can have a chance to edit them and keep repos a safe space.

It uses theTensorflow.js toxicity classification model.

It currently works when comments are posted on issues and PRs, as well as when pull request reviews are submitted.

If you want some details about how it works, feel free to check theblog post.

If you do not have any Github actions already set up in your repo, start by creating a .github/workflows folder.

Inside your workflows folder, create a new .yml file, for examplemain.yml and copy the following lines:

on:[issue_comment, pull_request_review]jobs:toxic_check:runs-on:ubuntu-latestname:Safe spacesteps:      -uses:actions/checkout@v2      -name:Safe space - action stepuses:charliegerard/safe-space@masterwith:GITHUB_TOKEN:${{ secrets.GITHUB_TOKEN }}

GITHUB_TOKEN isrequired (note that Githubautomatically creates this token) but two other parameters are optional:

• message - a custom message you'd like to display in the automatic comment
• toxicity_threshold - a float number between 0 and 1. It will be used when loading the machine learning model. Its default value is 0.9.

on:[issue_comment, pull_request_review]jobs:toxic_check:runs-on:ubuntu-latestname:Toxicity checksteps:      -uses:actions/checkout@v2      -name:Safe space - action stepuses:charliegerard/safe-space@masterwith:GITHUB_TOKEN:${{ secrets.GITHUB_TOKEN }}message:"this is my custom message"toxicity_threshold:0.7

The action can take up to 40s to run so, if you are testing it out in your repository, keep in mind that the bot will not display right after a toxic comment is posted.