Settings

Settings live in ~/.config/tuxaide/config.json. Change them with tuxaide config set <key> <value> — values are checked, and the change applies in the current terminal too — or go back with tuxaide config reset <key>. Reinstalling or updating keeps your values.

Answers

Key	Default	Meaning
model	picked for your RAM	The model that answers (models)
mode	llm	llm, smart or deep (Smart RAG)
temperature	0.1	0–2; low keeps answers focused
action_menu	true	After an answer, offer to put a command on the prompt or copy it
cache_ttl_days	30	Cached answers older than this are ignored (0–3650)
color	true	Colours in answers
system_context	true	Tell the model your OS, package manager, shell and init system
followup_window	600	Seconds a conversation stays open for follow-ups (0 turns them off)
followup_turns	3	Previous exchanges sent with a follow-up (0–10)

The shell

Key	Default	Meaning
enabled	true	Set by tuxaide on / off
failure_hint	true	After a failed command, show “type ? to ask TuxAide why”
typo_suggest	true	Suggest fixes for mistyped commands
session_capture	true	Keep the output of tuxaide run for the next question

Ollama and backends

Key	Default	Meaning
backend	ollama	ollama, or openai for an OpenAI-compatible API (remote backends)
ollama_url	http://localhost:11434	Where Ollama runs
api_base	—	Backend openai: the API address, e.g. http://localhost:1234/v1
api_key_env	OPENAI_API_KEY	Backend openai: the environment variable holding the key (the key itself is never stored)
prewarm	once (off under 8 GB)	Load the model when a terminal opens: off, once (at most every 10 min), always
keep_alive	10m	How long Ollama keeps the model in memory after a question (30s, 1h, 0, -1 = forever)
embed_model	nomic-embed-text	Embedding model for Smart RAG
rag_timeout	8	Seconds to wait for the man-page lookup before answering without it (1–120)

max_tokens, rag_top_k and rag_db_path are set by tuxaide mode; edit the file to change them by hand.