Ollama Cloud

Rho can use Ollama Cloud through its OpenAI-compatible API at https://ollama.com. This path does not need a local Ollama server for inference.

For local models, or cloud models routed through a signed-in local Ollama host, see Ollama.

At a glance

Setting	Value
Provider	ollama-cloud
Auth modes	ollama-cloud-api-key, ollama-cloud-device
Environment override	OLLAMA_API_KEY (API key mode)
Default API base	https://ollama.com/v1
Model list	Hosted catalog from /v1/models

One provider identity, two auth modes. Models stay ollama-cloud/<hosted-model> either way.

Setup with an API key

1. Create an API key at ollama.com/settings/keys.
2. In Rho, run:

/login ollama-cloud

Choose API Key and paste the key when prompted. Rho stores it in the configured credential store and refreshes the hosted model list.

3. Select a model:

/model ollama-cloud/<hosted-model>

for example /model ollama-cloud/kimi-k2.6 or /model ollama-cloud/kimi-k2.7-code.

Remove the stored key with:

/logout ollama-cloud

and choose the API key credential if more than one Ollama Cloud mode is stored.

Setup with a device key

Ollama's local install keeps an Ed25519 device key at ~/.ollama/id_ed25519. After that public key is registered with your ollama.com account, Rho can sign Cloud API requests the same way the Ollama CLI does.

1. Prefer the Rho login flow:

/login ollama-cloud

Choose Device Key. Rho opens the Ollama connect page (or prints the URL with --device-auth / headless login), waits until the device is approved, and stores a short session marker. Config keeps provider = "ollama-cloud" with auth = "ollama-cloud-device".

2. Or reuse an existing Ollama sign-in. If you already ran ollama signin and ~/.ollama/id_ed25519 is registered, Rho can use that key without creating an API key.

3. Select a model the same way:

/model ollama-cloud/<hosted-model>

/logout for the device mode clears Rho's session marker. It does not delete ~/.ollama/id_ed25519 or disconnect the key from ollama.com.

Environment and automation

For CI or development, set OLLAMA_API_KEY. It overrides a stored API key:

export OLLAMA_API_KEY="<api-key>"
rho --provider ollama-cloud \
  --auth ollama-cloud-api-key \
  --model ollama-cloud/kimi-k2.6 \
  run "review this project"

API keys are sent as Bearer tokens. Device-key auth signs each request with the local key and a ts query parameter; do not put either credential in config.toml or commit it to source control.

Device-key login from a remote shell:

rho login ollama-cloud --device-auth

When multiple login methods exist, prefer choosing Device Key in the TUI, or ensure the selected auth profile is ollama-cloud-device.

Models and compatibility

Use /config and choose Refresh model lists to fetch the current Ollama Cloud catalog. Hosted model ids can change as Ollama adds or retires cloud models; refresh before selecting a newly published model.

Ollama Cloud uses the same OpenAI-compatible chat path as local Ollama. Prefer models with reliable tool-call support for coding work. Rho does not claim that every hosted model supports tools, images, reasoning controls, parallel calls, or usage data.

If Ollama omits optional usage data, Rho still handles the response.