Features
Code Review

Code Reviewer

Automatically get reviews and feedback on code changes in your pull requests on GitHub.
Streamline code reviews and ensure high code quality commit messages across your codebase.

Code Review in Action

Key Features

  1. Automated Code Analysis: Analyzes code changes in pull requests and generates detailed feedback.

  2. Organized Feedback: Provides review comments organized by topics and confidence levels.

  3. Interactive Reviews: Allows engagement with the bot for clarification or additional context.

  4. PR Description Generation: Automatically generates descriptive summaries of code changes.

  5. Continuous Learning: Improves review quality over time based on user interactions.

How it Works

  1. Diff Analysis: When you create or update a pull request, the Kaizen Bot analyzes the code changes (diff) and generates detailed feedback based on the modified code snippets.

  2. Organized Feedback: The bot's feedback is organized into topics or categories like performance, security, code style, or documentation, making it easier to navigate and prioritize the comments.

  3. Confidence Levels: Each review comment includes a confidence level (critical, high, medium, low), indicating the perceived importance or severity of the issue.

  4. Contextual Information: The reviews provide context-specific details like file names, line numbers, code snippets, and explanations for the suggested changes.

  5. PR Description Generation: The bot can generate a descriptive summary of the code changes, helping you better document your pull requests.

Code Review Process

You can find an example here (opens in a new tab)

Usage Guide

  1. Getting Started: For a quickstart you can use the cloud hosted web application that we provide by following this. You can also choose to self host the Kaizen Github Bot locally by following the advanced guide here.
    The next given steps explore the latter approach in details.
  2. Create or Update a Pull Request: The Kaizen Bot will automatically analyze the code changes and generate a review.
  3. Review the Feedback: The bot's feedback will be shared as a comment on your pull request, organized by topics and confidence levels.
  4. Engage with the Bot: You can interact with the bot, provide additional context, or request clarification on its feedback.
  5. Iterate and Improve: As you work with the bot, it will learn from your responses and improve the quality of its reviews over time.

Example

Here's a complete example to review a pull request:

from kaizen.reviewer.code_review import CodeReviewer
from kaizen.generator.pr_description import PRDescriptionGenerator
from kaizen.llms.provider import LLMProvider
from github_app.github_helper.utils import get_diff_text, get_pr_files
from github_app.github_helper.pull_requests import clean_keys, create_review_comments
from kaizen.formatters.code_review_formatter import create_pr_review_text
import json
import logging
 
logging.basicConfig(level="DEBUG")
 
# Pull request information
pr_diff = "https://github.com/Cloud-Code-AI/kaizen/pull/335.patch"
pr_files = "https://api.github.com/repos/Cloud-Code-AI/kaizen/pulls/335/files"
pr_title = "feat: updated the prompt to provide solution"
 
# Get diff text and PR files
diff_text = get_diff_text(pr_diff, "")
pr_files = get_pr_files(pr_files, "")
 
# Create CodeReviewer instance
reviewer = CodeReviewer(llm_provider=LLMProvider())
 
# Review pull request
review_data = reviewer.review_pull_request(
    diff_text=diff_text,
    pull_request_title=pr_title,
    pull_request_desc="",
    pull_request_files=pr_files,
    reeval_response=False,
)
 
# Process review data
topics = clean_keys(review_data.topics, "important")
review_desc = create_pr_review_text(
    review_data.issues, code_quality=review_data.code_quality
)
comments, topics = create_review_comments(topics)
 
# Display results
print(f"Raw Topics:\n{json.dumps(topics, indent=2)}\n")
print(f"GENERATED REVIEW:\n{review_desc}")
print(f"\nComment and topics:\n{json.dumps(comments, indent=2)},\n{topics}")
 
# Generate PR description
print("---------------Generate desc-------------")
pr_desc = PRDescriptionGenerator(llm_provider=LLMProvider())
desc_data = pr_desc.generate_pull_request_desc(
    diff_text=None,
    pull_request_title=pr_title,
    pull_request_desc="",
    pull_request_files=pr_files,
    user="kaizen/example",
)
print(desc_data)

Supported Input

  • GitHub pull request information (diff, files, title, description)

Benefits

  • Improved Code Quality
  • Time Savings
  • Consistent Standards
  • Knowledge Sharing
  • Automated Documentation

Limitations

  • AI Limitations: While advanced, the bot may still have limitations in understanding complex code or context-specific nuances.
  • Human Oversight: The bot's feedback should be considered a complementary tool to human code reviews, not a complete replacement.
  • GitHub API Limitations: Rate limits and authentication requirements may affect large-scale usage.
Initial Setup GuideUnit Tests