Llama Coder

An open source Claude Artifacts – generate small apps with one prompt. Powered by Llama 3 on Together.ai.

Tech stack

• Llama 3.1 405B from Meta for the LLM
• Together AI for LLM inference
• esbuild-wasm + esm.sh for the in-browser preview renderer (runs in a sandboxed iframe)
• Next.js app router with Tailwind
• Braintrust for observability
• Plausible for website analytics

Cloning & running

1. Clone the repo: git clone https://github.com/Nutlope/llamacoder
2. Create a .env file and add your API keys:
   • Together AI API key: TOGETHER_API_KEY=<your_together_ai_api_key>
   • Database URL: Use Neon to set up your PostgreSQL database and add the Prisma connection string: DATABASE_URL=<your_database_url>
   • Braintrust API key (optional, for observability): BRAINTRUST_API_KEY=<your_braintrust_api_key>
   • Screenshot uploads: Configure S3_UPLOAD_KEY, S3_UPLOAD_SECRET, S3_UPLOAD_BUCKET, S3_UPLOAD_REGION, and IMAGE_UPLOAD_TOKEN_SECRET (a dedicated random secret, e.g. openssl rand -base64 32; it is required and must differ from S3_UPLOAD_SECRET). The uploader uses the existing bucket as-is. Uploaded screenshots are only needed for ~30 minutes (the validated-image token TTL), so add a bucket lifecycle rule that expires objects under next-s3-uploads/uploads/ after 1 day; this also removes uploads that were never verified. If the S3 credentials allow s3:DeleteObject, uploads that fail verification are deleted immediately.
3. Run npm install and npm run dev to install dependencies and run locally

Contributing

For contributing to the repo, please see the contributing guide