Add Ollama Cloud as built-in provider

[goemonwanwan]

Problem

Ollama Cloud (https://ollama.com) is a cloud inference service with OpenAI-compatible API. Currently, users must configure it via custom_providers, but this causes issues:

1. The provider: ollama-cloud setting is not recognized in auxiliary sections
2. Workaround requires provider: custom + redundant base_url + api_key in every auxiliary task config
3. PROVIDER_REGISTRY in auth.py does not include ollama-cloud

Proposed Solution

Add ollama-cloud to PROVIDER_REGISTRY in hermes_cli/auth.py:

"ollama-cloud": ProviderConfig(
    id="ollama-cloud",
    name="Ollama Cloud",
    auth_type="api_key",
    inference_base_url="https://ollama.com/v1",
    api_key_env_vars=("OLLAMA_CLOUD_API_KEY",),
),

Also add default auxiliary model in auxiliary_client.py:

_API_KEY_PROVIDER_AUX_MODELS["ollama-cloud"] = "nemotron-3-super"

Benefit

Before (workaround):

auxiliary:
  vision:
    provider: custom
    base_url: https://ollama.com/v1
    api_key: xxx
    model: qwen3.5:cloud

After (with built-in provider):

auxiliary:
  vision:
    provider: ollama-cloud
    model: qwen3.5:cloud

API key can be set via OLLAMA_CLOUD_API_KEY environment variable or custom_providers entry.

Related

• [Feature]: Local model routing for auxiliary tasks (Ollama/custom endpoint support) #879 (local Ollama routing) - related but different scope

[kshitijk4poor]

Addressed in #6038 — adds ollama-cloud as a built-in provider with dynamic model discovery (live API + models.dev, disk-cached).

[arsaboo]

@kshitijk4poor Will this also work with local Ollama? Local Ollama experience is not that great... cannot add them as fallback providers, etc.

[kshitijk4poor]

no sir, this is strictly for ollama-cloud for local model providers there should be a better way to use them is coming soon ;)

[arsaboo]

coming soon ;)

Cannot wait! If you needed any encouragement to work on it, you have some now 🙏

[Freffles]

Addressed in #6038 — adds ollama-cloud as a built-in provider with dynamic model discovery (live API + models.dev, disk-cached).

Sadly, looks like there were a couple of issues with that. Hope they can be resolved easily.

[kshitijk4poor]

Addressed in #6038 — adds ollama-cloud as a built-in provider with dynamic model discovery (live API + models.dev, disk-cached).

Sadly, looks like there were a couple of issues with that. Hope they can be resolved easily.

what's wrong with it? can you point out the issues?

[Freffles]

what's wrong with it? can you point out the issues?

There are two failing checks. Not sure if this is progressing.