OrcaRouter provider support for the Docker Model Runner gateway

Docker Model Runner makes running and serving LLMs feel native to the Docker workflow. You can pull a model and run docker model run ai/gemma3 "Hello" in one step, then reach the same model through a local OpenAI-compatible chat endpoint. The docker model gateway layers routing, fallbacks, and load balancing on top for teams that outgrow a single local model.

Because the gateway already configures hosted providers as provider/model entries with api_base and api_key, an optional OrcaRouter upstream would fit cleanly. Developers who want cloud-scale or frontier models, or a dependable fallback when local GPUs are busy, could add a hosted option next to openai and anthropic in the same YAML and keep using the same OpenAI-compatible client.

Proposal

I would like to propose OrcaRouter as an optional provider for the Docker Model Runner gateway. This would be purely additive: existing providers such as Docker Model Runner itself, Ollama, OpenAI, and Anthropic would remain untouched, and no user would need to configure OrcaRouter unless they want to.

OrcaRouter exposes an OpenAI-compatible API and uses standard API-key authentication. The expected integration point is the gateway's existing provider abstraction: a new entry whose model field points at OrcaRouter's OpenAI-compatible api_base, with the API key supplied the same way other provider keys are today (inline, os.environ/..., or ${VAR}). I have not implemented or tested anything against this codebase - this issue is to gauge maintainer interest first.

For Docker Model Runner users, the most relevant OrcaRouter capabilities are likely:

• Chat, reasoning, image, and video models from multiple vendors behind one endpoint, handy for evaluating a model before committing to it.
• Automatic model routing and provider failover, complementing the gateway's existing fallbacks for hosted traffic.
• Usage tracking and budgets, which matter once several developers share one gateway.

OrcaRouter is already used in open-source projects such as RAGFlow, Dify, and promptfoo, as well as by tools like goose, all over a standard OpenAI-compatible API.

Transparency note: OrcaRouter runs an optional open-source partner program in which approved OSS projects can receive a 5% revenue share from OrcaRouter usage attributed to their integration. Participation is not a prerequisite for integration, and I am happy to follow any disclosure or governance rules the project prefers. More context is available at https://www.orcarouter.ai/built-with.

I'm an engineer on the OrcaRouter team. Would the maintainers be open to an optional OrcaRouter provider? If so, I would be glad to submit an implementation PR following the project's conventions.

Start by reading the gateway's existing provider abstraction and the configuration conventions for Docker Model Runner, Ollama, OpenAI, and Anthropic. Done means an optional OrcaRouter provider can use its OpenAI-compatible api_base and the existing api_key forms without changing current providers.