Repolode
Submit
Nutlope

Nutlope/llamacoder

An open-source take on Claude Artifacts that generates small React apps from a single prompt, powered by Llama 3.1 405B via Together AI.

Open-source alternative to:Claude Artifacts

View on GitHubWebsite
Stars
7,156
Forks
1,676
Open issues
10
Last commit
16 days ago

What is it?

Llama Coder takes a natural-language prompt and produces a small web app, showing the generated code alongside a live preview. The LLM is Meta's Llama 3.1 405B, served through Together AI for inference.

The preview is rendered in the browser: esbuild-wasm compiles the generated code, and esm.sh supplies the imported packages. The output runs inside a sandboxed iframe. The app itself is a Next.js project using the app router and Tailwind, with Prisma connecting to a PostgreSQL database such as Neon.

Setup requires a Together AI key and a database URL. Braintrust observability and S3-based screenshot uploads are optional configuration, and the project is released under the MIT license.

Key features

  • 01Generates small apps from a single natural-language prompt.
  • 02Runs the generated code in a live in-browser preview inside a sandboxed iframe.
  • 03Uses Llama 3.1 405B for generation via Together AI inference.
  • 04Built on Next.js app router with Tailwind CSS and Prisma with PostgreSQL.
  • 05Optional Braintrust observability and S3 screenshot uploads.

Use cases

  • Quickly prototyping a small interactive UI from a text description.
  • Studying how an AI artifact generator is built with a browser-side renderer.
  • Self-hosting an AI app builder with your own Together AI key and database.

Quick start

  1. 1

    Clone the repo

    git clone https://github.com/Nutlope/llamacoder
  2. 2

    Create a .env file and add your API keys (Together AI, Neon database URL, optional Braintrust and S3 settings)

  3. 3

    Install dependencies and start the dev server

    npm install && npm run dev

Things to know

  • Requires a paid-tier Together AI API key and a PostgreSQL database (Neon is suggested) to run.
  • Screenshot uploads need an S3 bucket and extra secrets, adding setup effort for self-hosting.

Topics

#ai#llm#code-generation#llama#nextjs#react#typescript

FAQ

What is llamacoder?
An open-source take on Claude Artifacts that generates small React apps from a single prompt, powered by Llama 3.1 405B via Together AI. Llama Coder takes a natural-language prompt and produces a small web app, showing the generated code alongside a live preview. The LLM is Meta's Llama 3.1 405B, served through Together AI for inference.
Is llamacoder free to use? What license does it have?
llamacoder is open source under the MIT license. Check the license terms before using it in commercial products.
Is llamacoder still maintained?
Actively maintained. The last commit to llamacoder was on Oct 22, 2026.
What can I use llamacoder for?
Quickly prototyping a small interactive UI from a text description. Studying how an AI artifact generator is built with a browser-side renderer. Self-hosting an AI app builder with your own Together AI key and database.
How do I install llamacoder?
Clone the repo: git clone https://github.com/Nutlope/llamacoder Create a .env file and add your API keys (Together AI, Neon database URL, optional Braintrust and S3 settings) Install dependencies and start the dev server: npm install && npm run dev
Where can I get llamacoder?
The source code and installation instructions are on GitHub: https://github.com/Nutlope/llamacoder

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects