Ollama

Rho can use models served by Ollama through its OpenAI-compatible API. Ollama runs locally by default and needs no API key.

For direct hosted access with an API key or device key and no local server, see Ollama Cloud. If a local Ollama install is signed in with ollama signin, cloud models pulled as *-cloud still work through this local provider.

At a glance

Setting	Value
Provider	ollama
Auth	None
Default API base	http://127.0.0.1:11434/v1
Model list	Models installed in Ollama, refreshed from /v1/models

Setup

Install Ollama, start its server, and install a model that supports tool calls. For example:

ollama serve
ollama pull <tool-capable-model>

In Rho, open /config and choose Refresh model lists. The model picker then shows the models returned by Ollama. You can also select one directly:

/model ollama/<tool-capable-model>

Do not run /login for Ollama. Rho sends no Authorization header and does not read or write an Ollama credential.

Use another server

Set a provider-specific API base in ~/.rho/config.toml:

[providers.ollama]
base_url = "http://192.168.1.20:11434/v1"

Keep the /v1 suffix. Rho appends /models for discovery and /chat/completions for agent turns. The setting applies only to Ollama and is also used by /doctor when it checks the server. The URL must use http or https and cannot contain credentials, a query, or a fragment. Bearer tokens and custom headers for secured remote endpoints are not supported.

Model compatibility

Ollama's OpenAI-compatible endpoint can serve many model types, but not every model supports the tool calls needed by a coding agent. Choose a model with reliable tool-call support. Rho does not download models or claim that every installed model supports tools, images, reasoning controls, parallel calls, or usage data.

If Ollama omits optional usage data, Rho still handles the response. Run /doctor to distinguish a reachable server with installed models, a reachable server with no models, an unreachable server, and an invalid or unsuccessful response.