Refact.ai Updates for January 2024

In January 2024, we rolled out version 1.3 of the Refact.ai self-hosting server, introduced per-user statistics that help to understand how much AI helps to write code, and made a lot of updates to IDE plugins, including a new one for Visual Studio. Continue reading for more details!

Quickly Try Refact.ai Self-Hosting with Runpod.io

Now it’s easy to try our self-hosting server and the Enterprise version on Runpod.io. Here is the guide you can follow in a matter of minutes.

With our own GPU, you can try fine-tuning of a code completion model! It’s easy, no special skills required.

A fine-tuned model will adapt to your code base, coding style, and remember your APIs. As a result, you will get more relevant completions and increase the productivity of your team.

You can also try the Enterprise version of Refact.ai. It’s easy to set up: efficient inference, encrypted connection for multiple users, and per-user statistics.

Check out the video with Runpod.io deployment guide:

Introducing User Statistics

With the Refact.ai self-hosting server, you can now track how much code Refact.ai writes for your team. You can break down the stats by programming language, model, day or week. It’s great to see the impact on your productivity in numbers!

New Models

Added in January:

• Mistral 7B
• Mixtral 8x7B
• Deepseek 6.7B, 33B
• Magicoder 6.7B
• Stable Code 3B

The most important of these is Mixtral 8x7B. It’s the first time we see an open model working as good as GPT-models on code improvement tasks.

For example, you can try to make code shorter, find & fix bugs. Works in our in-IDE chat or with /shorter and /bugs commands in the AI Toolbox (press F1).

Running AI Toolbox /bugs command, using the Mistral 8x7B model

New Plugin: Visual Studio

Good news for Visual Studio Classic users, we’re out at your fav marketplace! Feel free to download and try it with this link.

It open source, too: https://github.com/smallcloudai/refact-vs-classic/

Better JetBrains Plugin

JetBrains IDEs are now faster, with many bugs fixed, and compatible with Android Studio. Get it at the JB marketplace!

Added AMD Support

Now, Refact.ai supports AMD cards. Check the details in this PR. As Refact.ai is open-source, this feature was contributed by our user. Many thanks, 88Ocelot 🙏

Self-hosted OSS and Enterprise Versions Updates

• Released 1.3 of Enterprise and OSS
• IDE plugins now immediately see changes in the models list
• Supports embedding models in anticipation of the upcoming RAG release
• Anthropic integration added
• The OSS version has login/password now, so it can be exposed to the internet
• Finetune fixes
• Enterprise Tag Upgrade: use latest tag to download docker images, instead of beta, because we are out of beta now! 🎉

Contribute to Refact.ai, an Open-Source AI Coding Assistant

Refact.ai is community-driven, if you want to contribute, discuss new features, join our Discord, and let’s shape the future of programming together!