Fallback model and provider

[JohnPreston]

Hello team,

I have configured different LLMs, and it would be great to have the ability to implement a chain of models and providers to use.

For example, let's say I have OpenRouter and my local LLM, I could have

  "llms_chains": {
    "default_code": [
      {
        "provider": "openrouter",
        "model": "openai/gpt-oss-20b"
      },
      {
        "provider": "local-llm",
        "model": "gpt-oss:20b"
      }
    ],
    "default_chat": [
      {
        "provider": "openrouter",
        "model": "gemma3"
      },
      {
        "provider": "local-llm",
        "model": "mistal-instruct"
      }
    ]
  }

Idea inspired of n8n AI Agent, the idea here is to have a series of fallback LLMs. If any error occur in the top-tier LLMs, move on to the next one, and so on.

Thanks for your consideration.

// I did my best to search in the issues and discussions and did not find a similar one