Install Ĝi

Five commands, then connect your model server.

1. Add the signed package source and install

sudo install -d -m 0755 /etc/apt/keyrings
curl -fsSL https://gi-ai.app/apt/gi-ai.asc | sudo gpg --dearmor -o /etc/apt/keyrings/gi-ai.gpg
gpg --show-keys /etc/apt/keyrings/gi-ai.gpg      # compare the fingerprint below
echo "deb [signed-by=/etc/apt/keyrings/gi-ai.gpg] https://gi-ai.app/apt stable main" \
  | sudo tee /etc/apt/sources.list.d/gi-ai.list
sudo apt update && sudo apt install gi-ai

2. Connect a model server

Ollama on the same computer (default)

ollama pull qwen3:4b
gi health

LM Studio on the Windows host (Ĝi in WSL)

Start the LM Studio server, load a model, and allow its port only from WSL in the Windows firewall. Then create ~/.config/gi-ai/gi.toml:

[llm]
backend = "lmstudio"
endpoint = "http://@gateway:1234"   # the Windows host as seen from WSL, found on every run
model = "qwen/qwen3.5-9b"           # the exact model id shown by LM Studio
allow_remote = true
timeout_s = 300

3. First steps

gi health
gi ask "Explain what a hash-chained audit log is in two sentences."
gi init-workspace && gi task new --type req --title "Try Ĝi"
gi selfcheck

Updates and removal

sudo apt update && sudo apt upgrade     # updates arrive with your normal updates
sudo apt remove gi-ai                      # your workspace in ~/.local/share/gi-ai stays yours