gaim

gaim

One aim from your gbrain

gaim asks your local gbrain, โ€œWhat should I do next?โ€, then gives you one source-backed aim.

View on GitHub gbrain

How it works

Pick one thing. Do it.

  1. 1

    Click the menu-bar icon. gaim asks your local gbrain what to do next.

  2. 2

    Local Ollama (gemma 4 4B MLX) picks one source-backed aim. No data leaves your machine.

  3. 3

    gaim opens a native dialog with the aim, why, and a 90-minute plan. Markdown session saved locally.

Local-first

No account. No dashboard.

gaim calls your local gbrain CLI and runs Ollama (gemma 4 4B by default) for AI reasoning. Session reports stay in ~/Library/Application Support/gaim/reports/. On first launch gaim asks once whether to use local Ollama or OpenRouter. OpenRouter is opt-in via the Settings dialog (right-click the tray icon); the API key sits in macOS Keychain.

Install

Paste, top to bottom.

# macOS, Apple Silicon recommended
brew install ollama node rust
ollama serve &                 # or open the Ollama app
ollama pull gemma4:e4b-mlx     # ~9.6 GB, MLX-tuned 4B

# gbrain CLI (required). Install per
# https://github.com/garrytan/gbrain

# Build gaim
git clone https://github.com/sliday/gaim gaim
cd gaim
npm install
npm run build

# Run
open src-tauri/target/release/bundle/macos/gaim.app

First launch asks Local (Ollama) vs OpenRouter. Pick Local for the private default.

Product taste

Small is the product.

The app has one decision button: What should I do next?. Explanation opens in a native scrollable modal. Session history stays in files.