Deep-Code: Open Source CLI Coding Agent

What is this?

A free, open-source alternative to proprietary AI coding assistants (like Claude Code). Deep Code is a CLI tool that lets you:

• Build, edit, and analyze code and apps using open-source LLMs (Groq API, DeepSeek, Llama, etc.)
• Automatically generate and save code files/folders for your requests—no more copy-paste!
• Use a privacy-first, community-driven, extensible coding agent.

Why?

• Proprietary tools are expensive and closed-source ($17–$200/month)
• You deserve full control, privacy, and transparency
• Open-source means no vendor lock-in and community-driven improvements

Progress So Far

• Project structure, MIT license, and CI/CD
• Python virtual environment support
• CLI with config, chat, and version commands
• Interactive config: set API key and select model (DeepSeek, Llama, etc.)
• Multi-turn coding chat: ask for an app, code is saved to files/folders automatically
• Clean output: only assistant replies, not raw API responses
• Error handling and user feedback

Improvements Still To Be Made

• Smarter folder and file naming (e.g., use app name, language, or user intent)
• Fewer commands: after config, launch coding chat automatically (no extra command needed)
• One-step startup: auto-create/activate venv and install dependencies if missing
• Global CLI install (e.g., just type ai-code anywhere)
• More robust multi-file project support
• Plugin system for extensibility
• Git integration, code analysis, and more

Supported Models

• deepseek-r1-distill-llama-70b
• meta-llama/llama-4-maverick-17b-128e-instruct
• meta-llama/llama-4-scout-17b-16e-instruct
• llama3-70b-8192
• llama-3.3-70b-versatile

Quick Start

1. Clone the repo and enter the directory:

   git clone <repo-url>
   cd ai-code

2. Create and activate a Python venv:

   python3 -m venv .venv
   source .venv/bin/activate
   pip install -r requirements.txt
   pip install python-slugify

3. Configure your API key and model:

   python3 src/cli/main.py config --set
   # (Select your model and enter your Groq API key)

4. Start coding! (multi-turn chat, code is saved automatically):

   python3 src/cli/main.py chat

   • Just type your request (e.g., "build me a calculator web app").
   • The agent will create a folder and files for your app automatically.
   • Type /exit to quit.

Tip:

• You can run all commands from your project root after activating the venv.
• If you see an error about slugify, run: pip install python-slugify

Roadmap

• Smarter file/folder naming
• Fewer startup steps
• Global CLI install
• Plugins, git, and more

License

MIT