🤖 Self-Improving Coding Agent MVP

A browser-based AI agent that generates, tests, and refines code using local LLM inference

Project Overview

A browser-based AI agent that generates code using local LLM inference (Ollama). Demonstrates real-time streaming, token tracking, and interactive help features.

✨ Key Features

🔄 Real-Time Streaming

Code appears as it's generated, providing immediate feedback

📊 Token Tracking

Monitor token usage with detailed breakdown (prompt + generated)

💡 Interactive Help

Contextual help buttons explain each component and file structure

🏥 Health Checks

Automatic dependency verification ensures everything works

🔍 Debug Panel

View raw prompts and responses for learning and debugging

🎭 Mock Mode

Test the UI without Ollama using simulated responses

🛠️ Tech Stack

React 18+ • Ollama (phi3:latest) • Monaco Editor • Vite

🚀 Quick Start

  1. ollama pull phi3:latest & ollama serve
  2. cd my-agent-app & npm install & npm run dev
Launch App View on GitHub Read Documentation

Video Demo

Watch the agent in action

Demo Video: Watch the Self-Improving Coding Agent in action. Try it yourself →

Try It Yourself