Our take
- Activity
- 38
- Community
- 53
- Issues
- 53
- Pull requests
- 43
+27 stars in the last 30 days
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
7,165Forks
1,672Change since Nov 16, 2026: +1 stars·-1 forks
Overview
Llama Coder is a web app for generating small applications from a single prompt. It uses Llama 3.1 405B through Together AI and renders previews in the browser, giving users a quick way to try an idea without building the first version by hand.
The repository is useful both as a working prompt-to-app tool and as a starting point for developers who want to study or adapt this kind of interface. Running it locally requires configuring the services and credentials it depends on.
Key Features
- Generates small apps from natural-language prompts.
- Uses Llama 3.1 405B for language-model responses.
- Calls Together AI for inference.
- Builds in-browser previews with esbuild-wasm and esm.sh in a sandboxed iframe.
- Uses Next.js App Router and Tailwind CSS.
- Supports optional Braintrust observability.
- Includes configuration for a PostgreSQL database and screenshot uploads.
Use Cases
- Developers can quickly prototype a small app idea from a prompt.
- Teams evaluating AI-generated interfaces can explore a working prompt-to-preview workflow.
- Contributors can use the codebase as a foundation for a customized app-generation web experience.
- Learners can inspect how model inference and sandboxed browser previews fit into a web app.
Getting Started
git clone https://github.com/Nutlope/llamacoder
cd llamacoder
npm install
npm run dev
Configure the required environment variables before running. See the README for API keys, database setup, upload configuration, and further details.
Considerations
- Local use depends on a Together AI API key and a PostgreSQL database URL; screenshot uploads require additional S3-related credentials.
- The project uses external inference rather than running its language model locally.
- Generated app previews run in a sandboxed iframe, but the repository description does not establish broader security guarantees for generated code.
- Braintrust is optional and only needed for the observability setup.
Found this useful?
Share it with someone who would like llamacoder.