Configuration

Pin both provider roles in mlola.config.json (working directory, then the runtime home) instead of exporting anything.

Overview#

Pin both provider roles in mlola.config.json (working directory, then the runtime home) instead of exporting anything.

mlola.config.json

{
  "providers": {
    "decision":  { "provider": "typesafe", "model": "jev-latest" },
    "reasoning": { "provider": "openai",   "model": "gpt-5.6-luna" }
  }
}

Role	Providers	What it decides
decision	typesafe \| openrouter \| keyword	the fast path: which operation and target to act on
reasoning	openrouter \| openai \| custom \| none	ESCALATE, planning and fallback; `planner` is an accepted alias

Each role also accepts model, apiKey, apiKeyEnv and baseUrl — that is how Azure-style endpoints and gateway keys are pinned.

Inspect it

mlola browser config init      # write the starter file (never overwrites)
mlola browser config show     # the file, plus what actually resolved
mlola browser config show --json