KO | EN

GitLite

Minimalist Open Source Repository Search Engine

gitlite — search

#dsh #dsh-plugin #ai-agents #developer-tools #codex #claude-code #mcp #rust #deepseek-harness #typescript #ai-agent #cli

chatgpt-action

★ 45 Open GitHub ↗

A collection of ChatGPT assistants, e.g., pull request scoring, code view, labeler, assigner, etc.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# ChatGPT actions

A collection of ChatGPT assistants, e.g., code viewer, labeler, assigner, etc.

- [Overview](#overview)
  - [Why reinvent the wheel?](#why-reinvent-the-wheel)
  - [Features](#features)
- [Usage](#usage)
  - [Configuration](#configuration)
    - [Environment variables](#environment-variables)
    - [Inputs](#inputs)
  - [Prompt templates](#prompt-templates)
    - [Variables available in prompt templates](#variables-available-in-prompt-templates)
- [Developing](#developing)
- [FAQs](#faqs)
  - [Review pull request from forks](#review-pull-request-from-forks)
  - [Choose the ChatGPT API implementation](#choose-the-chatgpt-api-implementation)
  - [Inspect the messages between ChatGPT server](#inspect-the-messages-between-chatgpt-server)
- [License](#license)

![https://github.com/unsafecoerce/chatgpt-action/pull/13](https://raw.githubusercontent.com/unsafecoerce/chatgpt-action/main/docs/images/chatgpt-example.jpg)

## Overview

### Why reinvent the wheel?

There're already many [chatgpt-actions][1], why we need to reinvent the wheel?

- Code review is not pull request changeset summarization

Some of existing actions send the whole pull request diff to ChatGPT, and
  ask ChatGPT to summarize the pull request, or summarize each hunks. In
  practice, the summarize has already done by the proposer and the code reviewer
  is responsible for reviewing the changes, evaluate the quality, and decide
  whether to merge the pull request.

**This action has a "score" action that asks ChatGPT to score the pull request,
  and give a short comment as a explanation. That should be a "good hint" when
  starting a code review.**

- Leave too much comments on Github pull requests

Some of existing actions send each hunks in a pull request diff to ChatGPT,
  and yields a comment (or comment snippet) for each hunk. That makes them in
  practically unusable, as receiving 20+ email notification after submitting
  a pull request is not a good experience. For the reviewer's perspective, we
  usually just focus on the hunks that are "too good" or "too bad".

**This actions take such experience into consideration, and only leave comments
  for the hunks that are not "LGTM".** (We do have an option to leave comments
  for good hunks as well).

- Lack of customization of the prompt

All of existing actions hard-coded the prompt template inside the source code
  and there's no chance to customize the prompt. When the built-in prompt is
  not good enough, ChatGPT responses with useless comments.

At the same time, someone may have good prompt ideas, he/she want to try using
  ChatGPT as an AI assistant for code quality evaluation and review, but he/she
  needs to reinvent the wheel for a new action to play on Github.

**This action makes the prompt template configurable.** With a set of variables
  like `$title`, `$description`, `$filename`, `$patch`, you can customize the
  message sent to ChatGPT for better experience.

Prompt engineering is an _art_ to make ChatGPT success, we encourage the community
  to try discovering new, novel, and insightful prompts to make ChatGPT do
  code review job better, _**without reinventing the wheel for another chatgpt-action**_.

**ANY SUGGESTIONS AND IMPROVEMENTS ARE HIGHLY APPRECIATED!!!**

### Features

- Scoring your pull requests

```yaml
      - uses: unsafecoerce/chatgpt-action@main
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          CHATGPT_ACCESS_TOKEN: ${{ secrets.CHATGPT_ACCESS_TOKEN }}
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
        with:
          debug: false
          action: score
  ```

- Code review your pull requests

- Customizable prompt templates

> $description

I would like to give you the whole diff of the pull request and you need to given a score after
            reviewing the pull request. The score should be a number between 0 and 100 and 85 means the
            the pull request is basically acceptable. You just need to reply a number between 0 and 100,
            e.g., "85", and a comment within 30 words. Reply "OK" to confirm.
  ```

## Usage

```yaml
name: Code Review

permissions:
  contents: read
  pull-requests: write

on:
  pull_request:

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
        with:
          repository: ${{github.event.pull_request.head.repo.full_name}}
          ref: ${{github.event.pull_request.head.ref}}
          submodules: false

- uses: unsafecoerce/chatgpt-action@main
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          CHATGPT_ACCESS_TOKEN: ${{ secrets.CHATGPT_ACCESS_TOKEN }}
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
        with:
          debug: false
          action: score

### Configuration

chatgpt-action

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README