cerebras_coder

This is an AI code assistant powered by Cerebras, running llama3.3-70b.

Inspired by Hassan's Llama Coder.

Architecture

The application has been converted to use Hono for better performance and caching:

• Backend: Hono-based API server (/backend/index.ts)
• Frontend: React SPA served from /public/
• Database: SQLite with migrations and queries
• Caching: HTTP caching headers for static assets and generated content

Setup

1. Sign up for Cerebras
2. Get a Cerebras API Key
3. Save it in your project env variable called CEREBRAS_API_KEY

Caching Strategy

• Static assets (JS, CSS, images): 24 hours cache
• Versioned content (/p/:project/v/:version): Immutable cache (1 year)
• Latest project (/p/:project): 5 minutes cache
• Main page: 5 minutes cache

API Endpoints

• GET / - Main application page
• POST /api/generate - Generate code from prompt
• GET /p/:project - Latest version of project
• GET /p/:project/v/:version - Specific version of project
• GET /public/* - Static assets