Local AI Gemma

A local-first VS Code extension that opens a right-side AI assistant panel and can call local tools such as reading files, writing files, listing folders, and running shell commands without relying on a remote AI service.

What is included

• VS Code webview panel on the right side
• Local Python backend for Gemma 2 9B inference from a local model path
• Tool hooks for workspace access and shell execution
• Optional extension installer command for local extension management

Offline setup

1. Place a local Gemma 2 9B model in a folder such as: models/gemma-2-9b-it
2. Set the environment variable: GEMMA_MODEL_PATH=/absolute/path/to/models/gemma-2-9b-it
3. Install Python dependencies: pip install -r python/requirements.txt
4. Launch the VS Code extension and use the Local AI panel.

If no local model is found, the backend returns a graceful fallback response and keeps the extension usable for local tool operations.

Notes

This project is designed to be local-first. It does not require internet access for normal interaction after the model and dependencies are installed.

The extension uses VS Code's APIs to access the workspace, terminal, and extension installation capabilities in a user-controlled way.