Support for Ollama Cloud API with API key authentication

[orrinwitt]

Problem Description

Currently, Nanobot's ollama provider is designed for local Ollama instances (no auth). Users can already access Ollama Cloud models by routing through a local Ollama instance, but this potentially adds unnecessary latency and requires running local infrastructure. There's no documented way to connect directly to Ollama's cloud API.

Ollama Cloud API Details

From Ollama's documentation:

• Endpoint: https://ollama.com/api
• Authentication: Bearer token in Authorization header
• Environment variable: OLLAMA_API_KEY

Proposed Solution

Add native support for Ollama Cloud API:

1. Support API key authentication in the ollama provider
2. Document configuration for both local and cloud Ollama
3. Handle Authorization header properly when apiKey is provided

Use Case

Users who want to:

• Test out an LLM completely for free through Ollama's limited cloud access before paying for it on something like OpenRouter
• Use Ollama's cloud-hosted models (e.g., gpt-oss:120b-cloud) without running local hardware
• Reduce latency by connecting directly to Ollama Cloud instead of routing through a local Ollama instance

Related Issues

• How Do I content the Ollama #75 - Local Ollama workarounds
• Ollama api support? #193 - Ollama API support discussion
• Proposal: Replace LiteLLM with native SDKs + enhanced local model support #161 - Proposal to replace LiteLLM with native SDKs

Documentation References

• Ollama Cloud Documentation
• Ollama API Authentication

---

Question: Has anyone successfully configured the ollama provider for Ollama Cloud? Would the maintainers prefer enhancing the existing ollama provider or creating a separate ollama-cloud provider?

[recluse90s]

I’m using the Ollama Cloud API as well (not local). Here’s a config that works for me—hope it helps :-)

"ollama": {
      "apiKey": "YOUR_OLLAMA_API_KEY",
      "apiBase": "https://ollama.com",
      "extraHeaders": null
}

[orrinwitt]

I’m using the Ollama Cloud API as well (not local). Here’s a config that works for me—hope it helps :-)

"ollama": {
      "apiKey": "YOUR_OLLAMA_API_KEY",
      "apiBase": "https://ollama.com",
      "extraHeaders": null
}

thanks, that does work! I'm still getting litellm transient errors with minimax-m2.7, but it does work with other models like glm-5, so this is the answer.