A note from a human

I originally wrote Vixl (originally named Pyrola) to create a nice-feeling agent harness UI to use Qwen on my Strix Halo.

Local LLMs helped me manage media on my local Jellyfin server, and test harnesses or projects.

At work I had used the Vercel AI SDK countless times to build features, so this project felt like a fun side project to see what it would take to give myself a "cloud feeling local LLM ux."

The resulting project is now my daily driver at work.

Like all personal projects, Vixl does have opinions. You can read them in Philosophy. I try to strictly limit them.

Start here

1. Install Vixl.
2. Set up a provider and a default model.
3. Add a project if you want the chat rooted in a folder. Pick two or more projects to create a project group. Home starts a home chat.
4. Send a first chat.