Configuration reference

Persistent settings live under ~/.xllama/. Environment variables override config.json, which overrides built-in defaults.

config.json

{
  "model": "gpt-4o-mini",
  "model_name": "gpt-4o-mini",
  "reasoning": "medium",
  "provider": "openai",
  "api_base": "",
  "api_version": "",
  "ctx": 32768
}

Key environment variables

Variable	Purpose
XLLAMA_MODEL	Default model spec
XLLAMA_PROVIDER	Provider id
XLLAMA_API_BASE	Override base URL
XLLAMA_API_KEY	Generic key fallback
XLLAMA_SERVER_URL	Local server URL
XLLAMA_LLAMA_BIN	Path to llama-server
XLLAMA_REASONING	low \| medium \| high
XLLAMA_CTX	Context window size
XLLAMA_MAX_TOKENS	Max response tokens
XLLAMA_BASH_TIMEOUT	Shell timeout; default 60s
XLLAMA_ALLOW_SUDO	Set to 1 to override the root-command guard
XLLAMA_NO_BG	Disable terminal background setting

CLI reference

xllama [--model <spec>]
xllama setup [--model …] [--provider …] [--api-key …]
xllama status
xllama install [--dir <path>]
xllama serve [--port N] [--model s] [--bg] [--stop]
xllama once [--plan] [--model s] <prompt>
xllama help