Refact.ai

Refact.ai AI Coding Assistant | Fine-tuned, Secure, and Open-source

Refact.ai is an open-source, self-hosted AI coding assistant that integrates directly into your development workflow. It combines retrieval-augmented generation (RAG) and fine-tuning to provide accurate code suggestions, in-IDE chat, and a customizable toolbox to explain, summarize, refactor, document, and debug code. It supports on-premise deployment, multiple large language models (LLMs), and broad language coverage, all while keeping data under your control.

---

How Refact.ai Works

• Integrates with your IDEs (VS Code, JetBrains, Visual Studio, Neovim, Sublime Text, PyCharm, WebStorm, GoLand, IntelliJ, CLion, and more).
• Uses Retrieval-Augmented Generation (RAG) plus fine-tuning to leverage your entire coding environment as context for accurate completions.
• Allows in-IDE chat: select code, ask for modifications, and push changes back as diffs.
• Lets you train or fine-tune models on your codebase to reflect your team’s style and practices.
• Provides a customizable Toolbox for code explanation, summarization, refactoring, documentation, and bug finding.
• Offers a choice of 20+ LLMs (e.g., Mistral, Llama3, GPT-4 for chat; Code Llama, StarCoder for completion) and an efficient Refact/1.6B model.
• Enables on-premise deployment for full data control; telemetry stays within your infrastructure. Open-source, with verifiable behavior.

---

How to Use Refact.ai

1. Install Refact.ai plugins for your IDE (VS Code, JetBrains, Sublime Text, Neovim, PyCharm, WebStorm, etc.).
2. Choose your preferred LLMs and configure your environment.
3. Fine-tune on your codebase if desired to improve relevance and adherence to internal guidelines.
4. Use the In-IDE Chat to ask questions, request code changes, or generate new code; apply diffs directly.
5. Leverage the Toolbox to explain, summarize, refactor, write documentation, or find bugs.

---

Core Features

• Self-hosted / on-premise deployment for complete data control
• Open-source foundation with transparent and auditable behavior
• In-IDE chat: ask questions, modify code, and apply changes as diffs
• Retrieval-Augmented Generation (RAG) plus fine-tuning for context-aware suggestions
• Customizable Toolbox for explanation, summarization, refactoring, documentation, and debugging
• Support for 25+ programming languages
• Support for 20+ LLMs (e.g., Mistral, Llama3, GPT-4, Code Llama, StarCoder) with a selectable best-fit model
• Fine-tuning on your company's codebase to improve style and practices
• Plug-and-play integration with popular IDEs: VS Code, JetBrains, Visual Studio, Neovim, Sublime Text, PyCharm, WebStorm, GoLand, IntelliJ, CLion, and more
• Telemetry and data governance designed to stay within your infrastructure

---

Why Choose Refact.ai

• Keeps sensitive code on your infrastructure without compromising AI capabilities.
• Open-source nature enables trust, security audits, and customization.
• Flexible model selection and fine-tuning align AI behavior with your team’s standards.
• Direct in-IDE workflow reduces context switching and accelerates development.

---

Safety and Security Considerations

• On-premise deployment ensures data never leaves your environment.
• Open-source code allows verification of model behavior and data handling.

---

Core Capabilities Summary

• Accurate code completion with context from your entire workspace
• In-IDE chat for real-time code modification and diffs
• Fine-tuning to reflect your company’s coding standards
• Broad language support and multiple LLM backends
• Fully self-hosted option with enterprise-grade security