🤖 LOVELY CODER – Agentic AI Software Engineer

LOVELY CODER is an end-to-end Agentic AI system that behaves like an automated software engineer. Inspired by tools such as Devin AI and Lovable, it converts a simple natural-language prompt into a fully functional, multi-file software project.

You describe what you want to build — Coder Buddy figures out how to build it.

✨ What Coder Does

Given a prompt like:

python main.py "Create a responsive calculator web application"

Coder will:

1. Plan the application features and tech stack
2. Design the project architecture and file structure
3. Generate clean, working source code (HTML, CSS, JS, etc.)
4. Iteratively refine the code until the project is complete

All generated files are written automatically to your local filesystem.

---

🚀 Key Features

• Agentic Planning
  Converts vague prompts into structured engineering plans and Jira-style feature stories.

• Architectural Reasoning
  Acts like a senior software architect by defining detailed implementation steps for every file.

• Iterative Code Generation
  Uses a ReAct-style agent loop to read, write, and refine code until completion.

• Tool-Enabled AI
  The agent can interact with the filesystem using controlled read/write tools.

• Visual Agent Debugging
  Supports tracing LangGraph execution and agent state using an Agentic Debugger.

• Schema-Driven Outputs
  Pydantic models enforce structured, production-grade outputs.

---

🧠 Agent Architecture

Coder Buddy is implemented using LangGraph with three core agent nodes:

1️⃣ Planner Agent

• Interprets the user prompt
• Defines:
  • Features
  • Tech stack
  • File structure
• Outputs a high-level engineering plan

2️⃣ Architect Agent

• Converts plans into detailed, file-level implementation instructions
• Acts like a senior engineer guiding development

3️⃣ Coder Agent

• Implements instructions
• Iteratively writes and refines code
• Continues until project status is marked done

---

🛠️ Tech Stack

Category	Technology
Agent Framework	LangChain, LangGraph
LLM Inference	Open-source GPT models via Groq Cloud
Language	Python
Package Manager	UV
Debugging	AI Agentic Debugger
Validation	Pydantic

---

▶️ Usage

python main.py "build a SIMPLE and COMPOUND INTEREST calculator web app with local storage"

▶️ Example Outputs:

⚠️ Disclaimer

This project is for educational and experimental use only.

👨‍💻 Developer Info

Name: Manjunath C Bagewadi
Email: manjunathsept11@gmail.com
Phone: +91 97428 96702
Porfolio: www.manjunathbagewadi.in
LinkedIn: linkedin.com/in/manjunath-bagewadi-9325ab55