🖥️ Local Coding Agent

A free, fully local AI coding agent platform generate Python code from natural-language instructions directly on your own computer.

Powered by Qwen2.5-Coder-7B.

✨ Highlights

• 🆓 Free to use — no costs, no LLM/API tokens or cloud inference charges

• 🔒 Local and Private — your prompts and generated code are made and stayed on your machine

• 🐍Specialized for Python code generation

• 🌐 Can also be used in Google Colab with a compatible Tesla T4 GPU

🏗️ How It Works

Type/prompt in the interface window, your program description. The coding agent will generate Python code, with code explanation. Then you can run the your program’s generated code separately.

💻 Requirements

Hardware

• A CUDA-enabled GPU, such as: RTX 4060 Ti (16GB), RTX 3090 / 4090 (24GB)

• CUDA-compatible GPU

• Recommended: **12 GB+ VRAM **for 4-bit quantization.

• System RAM: 16 GB+ recommended

• Storage: 20 GB+ recommended

Software

• Windows 10/11 or Linux

• Python 3.12

• PyTorch (Latest Stable), library torch

• CUDA-enabled PyTorch

• Transformers (Hugging Face): Specifically optimized to handle Qwen2.5-Coder-7B-Instruct

• accelerate

• bitsandbytes Essential for the 4-bit (NF4) quantization logic

📦 Run

python 

AI coding agent.py


On first run the the agent loads the Qwen coding model locally.

---------------------------------------------------------------------
  
 A private, free AI coding assistant running on your own GPU. 

 Software Created by Erez-Yahalomi ארז-יהלומי