Refact.ai Updates for July 2024
Summer’s in full swing, and we’ve got some hot updates in Refact.ai to match! In July, we made the chat AI more autonomous with your context, added GPT-4 mini, brought new models to our Enterprise and Self-hosted users, and more.
But before diving into all that—let us announce something big.
🎉 AI Agent by Refact.ai: Get early access
Imagine an AI that writes, debugs, and deploys code. Autonomously. For you.
Yes, we’re developing an AI coding agent in Refact.ai!
This isn’t just an upgrade - it’s how we’ll all code tomorrow. Soon, our AI agent will be able to run code in an isolated environment, use test-time computing, form a knowledge database about your company, and more.
Want to be the first to try? We are looking for 5-10 adventurous programmers to apply our AI agent in the real world, and share use cases and feedback.
Apply for Refact.ai AI Agent early access:
- Get the exclusive version of our AI Agent
- Plus, a special API plan with extra context to try it at full capacity.
To join, fill out this quick form and we’ll get in touch via Discord.
Tool Use: More Autonomous Chat AI
Now, when you ask questions about your code, AI can autonomously find and use relevant context from your project. No more manual @commands are needed—you get more accurate and contextual responses without any extra effort on your part!
Tool Use is a feature in Refact.ai’s in-IDE chat. Update your plugin and turn it on when chatting about your codebase. For best results, enable AST and VecDB in Settings.
You can read more about Tool Use and the best LLMs to use with it in our blog: Introducing Tool Use.
New Model: GPT-4o Mini
We’ve added support for GPT-4o Mini in Refact.ai’s chat—the latest model from OpenAI! It is available as a chat option for users with individual plans.
GPT-4o Mini is faster and smarter than GPT-3.5 Turbo, the previous base industry standard. It has more recent training data and shows better coding skills.
Try it in the Refact.ai chat to assist with various coding tasks: asking questions, debugging, generating code, creating documentation, unit tests, and more. It offers:
- 8K tokens for Free users
- 32K tokens for Pro users.
Latest Models Support in Visual Studio
Refact.ai plugin for Visual Studio now supports newer models like Starcoder2, Llama3, and others.
VecDB File Limit
The new setting regulates how effectively AI handles larger projects. You can increase it to improve the processing of multi-file codebases in your VS Code plugin.
Enterprise and Self-hosted Updates
Claude-3.5:
- Available in Docker for both Enterprise and Self-hosted users
- Requires an Anthropic API key
Llama3/8b/instruct:
- Available for both Self-Hosted and Enterprise users
- Enterprise users get access to the vLLM version for improved performance
Refact.ai: An Open-source Copilot Alternative with Customization and Advanced AI Coding Features
That’s all for our July updates. We hope you find these improvements useful!
- Try Refact.ai Pro: Check out our Pricing page for 1 month free.
- Get Refact.ai for your dev team: Set up a quick demo call with us and get a 2-week trial of Refact.ai Enterprise
We’d love your feedback:
- Join our Discord to discuss new functions or share feedback
- Check Refact.ai GitHub for our source code — Refact.ai is an open-source AI coding assistant and welcomes your contributions!
- Please rate us on VS Code and JetBrains
- And stay (fine-)tuned for further updates!