DevToolsAi PR Reviewer

This project is inspired by CodeRabbit. However, the original CodeRabbit is no longer maintained, so it was forked, improved, and rewritten from scratch in Python. Another motivation was to use it within Stellantis and run it on internal deployed models, ensuring that no data is sent to OpenAI servers.

AI-based PR reviewer and summarizer

Overview

DevTools pr-reviewer-ai is an AI-based code reviewer and summarizer for GitHub pull requests using Mistral's large and small models, deployed on a custom cluster or in Azure.It is designed to be used as a GitHub Action and can be configured to run on every pull request and review comments.

Reviewer Features:

• PR Summarization: It generates a summary and release notes of the changes in the pull request.
• Line-by-line code change suggestions: Reviews the changes line by line and provides code change suggestions.
• Continuous, incremental reviews: Reviews are performed on each commit within a pull request, rather than a one-time review on the entire pull request.
• Cost-effective and reduced noise: Incremental reviews save on OpenAI costs and reduce noise by tracking changed files between commits and the base of the pull request.
• "Light" model for summary: Designed to be used with a "light" summarization model (e.g. mistral-small) and a "heavy" review model (e.g. mistral-large). For best results, use mistral-large as the "heavy" model, as thorough code review needs strong reasoning abilities.
• Chat with bot: Supports conversation with the bot in the context of lines of code or entire files, useful for providing context, generating test cases, and reducing code complexity.
• Smart review skipping: By default, skips in-depth review for simple changes (e.g. typo fixes) and when changes look good for the most part. It can be disabled by setting review_simple_changes and review_comment_lgtm to true.
• Customizable prompts: Tailor the system_message, summarize, and summarize_release_notes prompts to focus on specific aspects of the review process or even change the review objective.

To use this tool, you need to add the provided YAML file to your repository and configure the required environment variables, such as GITHUB_TOKEN (could be done automatically with create-github-app-token). For more information on usage, examples, contributing, and FAQs, you can refer to the sections below.

• Overview
• Reviewer Features
• Install instructions
• Conversation with DevToolsAi
• Examples
• Contribute

Install instructions

pr-reviewer-ai runs as a GitHub Action. Add the below file to your repository at .github/workflows/pr-reviewer-ai.yml

name: Code Review

permissions:
  contents: read
  pull-requests: write

on:
  schedule:
    - cron: '0 1 * * *'  # Runs at 1 AM UTC every day
  workflow_dispatch:
  pull_request:
  pull_request_review_comment:
    types: [created]

concurrency:
  group:
    ${{ github.repository }}-${{ github.event.number || github.head_ref ||
    github.sha }}-${{ github.workflow }}-${{ github.event_name ==
    'pull_request_review_comment' && 'pr_comment' || 'pr' }}
  cancel-in-progress: ${{ github.event_name != 'pull_request_review_comment' }}

jobs:
  review:
    runs-on: ${{ vars.TAD_RUNNER }}
    timeout-minutes: 15
    steps:
      - name: "Create GitHub App Token"
        id: create-github-app-token
        uses: actions/[email protected]
        with:
          app-id: ${{ vars.GH_ACTION_SSH_APP_ID }}
          private-key: ${{ secrets.GH_ACTION_SSH_APP_PRIVATE_KEY }}
          owner: ${{ github.repository_owner }}

      - name: "Action Setup Python"
        uses: actions/setup-python@v4
        with:
          python-version: '3.11'

      - name: "AI PR Reviewer"
        uses: Stellantis-ADX/pr-reviewer-ai@master
        env:
          GITHUB_TOKEN: ${{ steps.create-github-app-token.outputs.token }}
        with:
          light_model_token_azure: 'token'
          heavy_model_token_azure: 'token'
          allow_empty_review: 'false'
          less_spammy: 'false'
          debug: 'false'
          review_simple_changes: 'false'
          review_comment_lgtm: 'false'

Environment variables

• GITHUB_TOKEN: This should already be available to the GitHub Action environment. This is used to add comments to the pull request.

Models: mistral-small and mistral-large

Recommend using mistral-small for lighter tasks such as summarizing the changes (mistral_light_model in configuration) and mistral-large for more complex review and commenting tasks (mistral_heavy_model in configuration).

Costs: as all our models runs internally, there is no cost associated with the usage of the models.

Prompts & Configuration

See: action.yml

Tip: You can change the bot personality by configuring the system_message value. For example, to review docs/blog posts, you can use the following prompt:

Tip: Once the Github action is deployed, you need to trigger workflow_dispatch to run the cache generation. It will speed up the action at least twice.

Blog Reviewer Prompt

system_message: |
  You are `@devtoolsai` (aka `github-actions[bot]`), a language model. 
  Your purpose is to act as a highly experienced
  DevRel (developer relations) professional with focus on cloud-native
  infrastructure.

  Company context -
  CodeRabbit is an AI-powered Code reviewer.It boosts code quality and cuts manual effort. Offers context-aware, line-by-line feedback, highlights critical changes,
  enables bot interaction, and lets you commit suggestions directly from GitHub.

  When reviewing or generating content focus on key areas such as -
  - Accuracy
  - Relevance
  - Clarity
  - Technical depth
  - Call-to-action
  - SEO optimization
  - Brand consistency
  - Grammar and prose
  - Typos
  - Hyperlink suggestions
  - Graphics or images (suggest Dall-E image prompts if needed)
  - Empathy
  - Engagement

Conversation with DevToolsAi

You can reply to a review comment made by this action and get a response based on the diff context. Additionally, you can invite the bot to a conversation by tagging it in the comment (@devtoolsai).

Example:

@devtoolsai Please provide a refactoring of this file as a senior developer.

Note: A review comment is a comment made on a diff or a file in the pull request.

Ignoring PRs

Sometimes it is useful to ignore a PR. For example, if you are using this action to review documentation, you can ignore PRs that only change the documentation. To ignore a PR, add the following keyword in the PR description:

@devtoolsai: ignore

Less spammy option

To reduce the number of comments made by the bot, you can set the less_spammy. It will keep only comments that involve a conversation with the user. It will delete all unresolved comments before doing the next review.

with:
    less_spammy: 'true'

Examples

Some of the reviews done by pr-reviewer-ai

Any suggestions or pull requests for improving the prompts are highly appreciated.

Contribute

Developing

First, you'll need to have at least python 3.11. Secondly, the basic understanding of poetry.

Install the dependencies

$ poetry install

Disclaimer

• Your code (files, diff, PR title/description) won't be shared with OpenAI.
• For the Stellantis users, all the models are deployed internally, so no data is sent to OpenAI servers.