# Configuration

## super-ollama config file

Resolved in order:

1. `$XDG_CONFIG_HOME/super-ollama/config.toml` when `XDG_CONFIG_HOME` is set
2. Otherwise `~/.super-ollama/config.toml`

Fields (TOML):

| Field | Purpose |
|-------|---------|
| `default_model` | Model name when `--model` is omitted (built-in default: `gemma3:1b`) |

## Logging

| Behavior | Condition |
|----------|-----------|
| WARN and above on stderr | Default (quiet inference logs) |
| Same levels as upstream `envconfig.LogLevel()` | `OLLAMA_DEBUG` set to any non-empty value |

## Model files

Models live in the usual Ollama layout on disk. Use the **`ollama`** binary from this repo (or another install) for **`ollama pull <model>`**.

## Environment variables

Inference and hardware behavior follow **Ollama** env vars (`OLLAMA_*`) where the shared engine reads them. See upstream documentation for GPU, context, and debug options.