Using Ĝi
The full reference is in the manual: man gi.
Commands
| Command | What it does |
|---|---|
gi health | Checks Ĝi and the model server. Exit 0 when the server answers correctly. |
gi ask "…" | Asks the model a question. --verbose adds speed statistics. |
gi init-workspace | Creates your private workspace (folder mode 0700) with a tamper-evident log. |
gi task new --type req|bug|vuln --title "…" | Creates a private task record; gi task list lists them. |
gi selfcheck | Checks your security baseline: private workspace, intact log, safe config, private model address. |
Add --json for machine-readable output.
Settings
System defaults are in /etc/gi-ai/gi.toml; your own settings go in ~/.config/gi-ai/gi.toml.
Key ([llm]) | Meaning |
|---|---|
backend | ollama (default), lmstudio, or echo (offline test) |
endpoint | http://host:port; @gateway means the Windows host seen from WSL |
model | The model name or id your server offers |
allow_remote | Must be true for any address that is not on this computer |
timeout_s | Seconds to wait for an answer (1–3600) |
reasoning | LM Studio only: off (default), low, medium, high, on, default |
token_file | LM Studio API token file, inside ~/.config/gi-ai/, mode 0600 |