> ## Documentation Index
> Fetch the complete documentation index at: https://docs.keinsaas.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Local Models

> Connect Navigator to Ollama or another local OpenAI-compatible model server.

Navigator can work with local or self-hosted model endpoints when they expose a compatible API. This is useful for private workloads, offline experiments, and avoiding per-token usage on external providers.

## Ollama

Start Ollama locally, pull a model, and point Navigator at the endpoint.

```bash theme={null}
ollama pull llama3.1
ollama serve
```

```dotenv theme={null}
OLLAMA_BASE_URL=http://localhost:11434/api
```

## OpenAI-compatible providers

Use OpenRouter or another compatible gateway when you want model choice through a common API style.

```dotenv theme={null}
OPENROUTER_API_KEY=
```

## Checks

* The local server is reachable from the Navigator runtime
* The selected model is downloaded or available from the provider
* The base URL includes the path expected by the integration
* Hosted deployments can reach the model endpoint if the endpoint is not local-only

<Info>
  A browser running at `app.example.com` cannot call `localhost` on your laptop through the hosted server. Use local development, a tunnel, or a reachable model gateway for hosted deployments.
</Info>