Sourcery is an AI-powered tool designed to automate code reviews and act as an IDE coding assistant. It aims to improve code quality, accelerate development velocity, and reduce the time spent on manual code reviews for developers and teams.

How It Works

Sourcery leverages large language models (LLMs) from OpenAI and Anthropic to analyze code changes within pull requests and provide instant feedback. It generates summaries, high-level comments, and line-by-line suggestions, mimicking a human code reviewer. As an IDE assistant, it offers real-time suggestions for code understanding, test generation, optimization, and quality improvements.

Quick Start & Requirements

• Code Reviews: Add Sourcery to GitHub repositories. Free for public/open-source projects; 14-day free trial for private repos, then requires a paid plan.
• IDE Assistant: Install via editor-specific plugins (PyCharm, VS Code, Sublime, Vim).
• Dependencies: Requires integration with GitHub for pull request reviews. LLM usage involves sending code diffs to OpenAI and Anthropic.

Highlighted Details

• Automated code reviews on GitHub pull requests.
• IDE assistant for real-time coding suggestions and refactoring.
• Utilizes OpenAI and Anthropic LLMs for review generation.
• Offers features like unit test and docstring generation.

Maintenance & Community

• Feedback can be provided via email to hello@sourcery.ai.
• Links to the Sourcery website and editor plugins are available.

Licensing & Compatibility

• Free for public repositories.
• Requires paid plans for private repository usage after a trial period.
• Code is sent to third-party LLM providers (OpenAI, Anthropic) for processing.

Limitations & Caveats

Code snippets are sent to third-party LLM providers for review generation. While Sourcery states code is not stored for more than 30 days and not used for training, users should be aware of this data transmission for private code.