Skip to content

Latest commit

 

History

History
187 lines (122 loc) · 8.19 KB

README.md

File metadata and controls

187 lines (122 loc) · 8.19 KB
Raw

Markdown Bot: an AI friend who can improve your Markdown content

A content suggestion robot powered by ChatGPT

This bot will make GitHub suggestion comments to update the contents of a file in a PR

When a user comments ai fix: <filename> or ai fix: <file1>, <file2> the bot will respond with inline suggestions on the PR.

A prompt can also be added by adding prompt: <prompt> after the filename to the comment.

For example:

ai fix: content/blog/Supercharge-Your-Markdown-Blog-With-AI.md
prompt: only return a signle suggestion

Example

Table of Contents

Implementation Overview

This bot can be implemented in two ways:

  1. As a GitHub App (self-hosted)
  2. As a GitHub Action

The Github Action implementation is simpler and easier to get started but it will require a github action to run on every comment to the repo. This could have a large impact your GitHub Action minutes if your repo has a lot of comments. The GitHub App implementation is more complex to setup. The Github App will run on a separate server and will not impact your github action minutes.

A GitHub App can be hosted on any hosting platform that supports Node.js. It can also be hosted in a serverless environment such as AWS Lambda or Netlify Functions.

NOTE: Serverless functions sometimes have timeout limitations or stop after a request responds with a 202 status code (This is the behavior in Vercel). This may cause the bot to fail if the OpenAI API takes too long to respond.

GitHub App Implementation

We do not provide a hosted version of the GitHub App so you will need to self-host it. You can self-host it on most hosting platforms. We have set up and tested it on Netlify (Functions) and Render.

Create a GitHub App

Follow the steps below or watch the video

Creating a GitHub App

  1. Go to your GitHub settings, create a new GitHub App
  2. Fill in the details (can put in temp values for the URL and webhook URL. We can update these later)
  3. Under permissions select Read for Contents. Select Read & Write for Pull Requests and Issues.
  4. Under "Subscribe to Events" select Issue comment
  5. Generate a webhook secret and copy it into the Webhook Secret field and save it somewhere safe
  6. Make sure "Only on this account" is selected under "Where can this GitHub App be installed?"
  7. Click "Create GitHub App"
  8. In the app settings click on "Generate a private key" and save the .pem file somewhere safe

Hosting the Bot Backend

For the GitHub App implementation of this bot, you need to host the backend somewhere. This sets up an API that will be requested once a user on your repo comments on a line in your PRs.

Self-hosting on Netlify

NOTE: functions are limited to a 10 second timeout on Netlify. This may cause the bot to fail if the OpenAI API takes too long to respond. To avoid this a background function can be used instead of the normal function. Netlify background functions are only available on the paid tier.

Self-hosting on Netlify

Deploy to Netlify

This will clone the repo to your GitHub account and deploy it to Netlify. See below for the required env variables.

Self-hosting on Render

NOTE: when using the free tier you may get frequent 503 errors due to the app being put to sleep. To avoid this you can upgrade to the paid tier.

Self-hosting on Render

Deploy to Render

ENV Variables

When setting the bot backend on your hosting provider, you will need to set the following environment variables:

# The GitHub App ID
APP_ID=***

# The Client ID of the GitHub App
GITHUB_CLIENT_ID=***

# The Client Secret of the GitHub App
GITHUB_CLIENT_SECRET=***

# The private key of the GitHub App (Copy and pasted from the .pem file)
PRIVATE_KEY=***

# The webhook secret of the GitHub App (User generated and pasted into the GitHub App)
# Can be generated by running: `ruby -rsecurerandom -e 'puts SecureRandom.hex(20)`
WEBHOOK_SECRET=***

See the ProBot docs & ProBot Configuration Guide for more details.

Install the GitHub App

Finally you can install the Github App on your repo. To do this:

  1. Go to the GitHub App settings
  2. Click on "Install App" Tab
  3. Click "Install" next to the account you want to install it on
  4. Select the repo(s) you want to install it on
  5. Click "Save"

GitHub Action Implementation

When implementing this bot as a GitHub Action it will run on every comment to every pull request and issue on the entire repo. This could be an issue if your repo has a significant amount of comments. The bot will only respond to comments on a PR and only if the comment follows the format ai fix: <filename> or ai fix: <file1>, <file2>.

Add the GitHub Secret

  1. Go to the repo homepage in which you want to integrate this bot
  2. Click "Settings"
  3. Click "Actions" under "Secrets and Variables"
  4. Change to the "Secrets" tab, create a new OPENAI_API_KEY secret with the value of your OpenAI API key

Add The GitHub Action to your repo

.github/actions/content-suggestion.yml

name: Content Suggestion

permissions:
  contents: read
  pull-requests: write

on:
  issue_comment:
    types: [created]
jobs:
  make-suggestion:
    name: "Make Suggestion"
    runs-on: ubuntu-latest
    steps:
      - uses: tinacms/ai-content@v0
        with:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

Testing The GitHub Action

The GitHub Action will run on any comment to a PR/issue, but it will only respond to comments on a PR if the comment follows the format ai fix: <filename> or ai fix: <file1>, <file2>.

To try it out: On a PR, comment ai fix: <filename>.

Contributing / Running Locally

If you want to run locally or if you have suggestions for how the content suggestion bot could be improved or want to report a bug, open an issue! We welcome all contributions.

For more information, please refer to the Contributing Guide.

Credit

This project is inspired by ChatGPT-CodeReview.

Contributing

If you have suggestions for how this project could be improved, or want to report a bug, open an issue! We'd love all and any contributions.

For more, check out the Contributing Guide.

License

MIT