Code Genius

AI-powered CLI developer assistant — 100% local, powered by Ollama

Code Genius leverages locally hosted Large Language Models through Ollama to provide intelligent code analysis, repository understanding, code review, documentation generation, and developer productivity features directly from the terminal.

Your code never leaves your machine. No API keys. No cloud dependencies. No data transmission.

⚡ Quick Start

# Install globally
npm install -g code-genius

# Initialize in your project
cd your-project
code-genius init

The init wizard will:

• Detect your operating system
• Verify Ollama installation
• Install required models (Qwen Coder + Nomic Embed)
• Generate local configuration
• Validate everything works

📋 Commands

Code Analysis

| Command | Description | |---|---| | code-genius review <file> | AI-powered code review with focus areas | | code-genius explain <file> | Detailed code explanation | | code-genius refactor <file> | Refactoring suggestions | | code-genius analyze | Project structure & architecture analysis |

Code Generation

| Command | Description | |---|---| | code-genius test <file> | Generate unit tests | | code-genius doc <file> | Generate documentation | | code-genius commit-msg | Generate commit messages from staged changes |

Repository Intelligence (RAG)

| Command | Description | |---|---| | code-genius index | Index your repository for semantic search | | code-genius ask "question" | Ask questions about your codebase | | code-genius chat | Interactive AI chat session |

Configuration

| Command | Description | |---|---| | code-genius init | Setup wizard | | code-genius config list | Show current configuration | | code-genius config set <key> <value> | Update a setting |

🔍 Examples

# Review code for security issues
code-genius review src/auth.js --focus security

# Explain a specific function
code-genius explain src/utils.js --start 10 --end 50

# Generate tests with Vitest
code-genius test src/helper.js --framework vitest --output tests/helper.test.js

# Generate a commit message (Conventional Commits)
code-genius commit-msg --conventional

# Index your repo and ask questions
code-genius index
code-genius ask "how does the authentication flow work?"

# Start interactive chat with RAG
code-genius chat --rag

⚙️ Configuration

Configuration is stored at two levels:

• Global: ~/.config/code-genius/config.json
• Project: .code-genius/config.json (project overrides global)

Key settings:

| Key | Default | Description | |---|---|---| | model | qwen2.5-coder:7b | LLM for code tasks | | embeddingModel | nomic-embed-text | Embedding model | | ollamaUrl | http://localhost:11434 | Ollama server URL | | temperature | 0.3 | LLM temperature | | topK | 5 | RAG retrieval results |

🏗️ Architecture

Developer → Code Genius CLI → Ollama Local API → LLM
                  ↓
         Repository Scanner → Chunker → Embeddings → Vector Index → RAG

🔒 Security & Privacy

• ✅ 100% local execution
• ✅ No source code transmission
• ✅ No API keys required
• ✅ No cloud dependencies
• ✅ Suitable for proprietary codebases

📦 Requirements

• Node.js 18+
• Ollama installed and running (ollama serve)
• ~5GB disk space for default models

License

MIT