Remote backends

TuxAide is local by default and stays that way unless you change it. If your machine is too small for a local model, or you already run a model server, you can point TuxAide at it.

Whenever a question leaves your computer, you can see it: ☁ Asking <host> while it waits, ☁ remote · <host> in every answer header, and a warning in tuxaide status and tuxaide doctor. Pre-loading the model at shell start only ever talks to a local Ollama.

Ollama on another computer

# on the other computer (it must listen on the network):
OLLAMA_HOST=0.0.0.0 ollama serve

# on this one:
tuxaide config set ollama_url http://192.168.1.10:11434
tuxaide model                 # pick one of its models

Any OpenAI-compatible API

LM Studio, llama.cpp server, vLLM, or a cloud service — anything with /v1/chat/completions:

tuxaide config set backend openai
tuxaide config set api_base http://localhost:1234/v1     # e.g. LM Studio on this machine: still local
tuxaide model                                           # the models the API offers

API keys

A service that needs a key reads it from an environment variable — TuxAide never writes it to any file, refuses tuxaide config set api_key, and tuxaide doctor flags a key found in config.json:

echo 'export OPENAI_API_KEY=sk-…' >> ~/.zshrc
tuxaide config set api_key_env MY_KEY_VAR     # if your variable has another name

Error messages never show the key.

Small machines

On a machine with less than 3 GB of RAM, the installer offers three choices: a tiny local model (qwen2.5:0.5b, 398 MB), Ollama on another computer, or an OpenAI-compatible API. Remote is never chosen for you.

Smart RAG with a remote backend

Smart RAG computes embeddings with Ollama (ollama_url). With the openai backend it still uses ollama_url for them — if that is remote too, the ☁ indicator says so.

Back to local

tuxaide config reset backend
tuxaide config reset ollama_url