Skip to main content
Muna is the easiest way to run arbitrary models from Huggingface.

​
Using our Compiler Agent in Slack

Simply join our Slack and ask Cursor to compile a model, providing a link to the Huggingface repo:
Asking our agent to compile a HF model.
Our compiler agent is powered by Agent Skills that provide detailed instructions on how to write and compile Huggingface models.
The agent will spend anywhere from a few minutes to a few hours writing the required Python inference code for the model, compiling it, and testing that it works:
HF model successfully compiled by our agent.

​
Running the Compiled Function

Once compiled, you can run the compiled model anywhere: 
import { Muna } from "muna"

// 💥 Create your Muna client
const openai = new Muna({ accessKey: "..." }).beta.openai;

// 🔥 Make a prediction
const transcription = await openai.audio.transcriptions.create({
    model: "moonshine/moonshine-base",
    file: audioFile
});

// 🚀 Use the results
console.log(transcription.text);

​
Try it Yourself

Compile in Slack

Compile a Huggingface model in our community Slack.

View Compiled Models

Check out some models that have been compiled by our agent.