Native Support for Mozilla.ai's any-llm

[angpt]

Hi Camel Team and Community,

I'm the DevRel Lead at Mozilla.ai. I have been following Camel AI, and I love the work you're doing with communicative agents.

I'd like to propose integration any-llm as a supported backend in Camel.

any-llm is a Python library by Mozilla.ai that provides a unified interface for LLMs. It allows developers to swap between OpenAI, Anthropic, Gemini, and Local Models (via Ollama/LocalAI) with a single configuration change, and without leaking data. Integrating with any-llm will reduce the need for Camel to maintain separate adapters for every new model provider that pops up.

I am happy to submit a PR for this. My plan is to implement a wrapper class inheriting for BaseModelBackend that routes requests through any-llm. This would allow users to simply pass an any-llm instance when initializing agents.

Is this an integration the core team would be open to merging? Looking forward to your thoughts.

Best
Anushri

[Wendong-Fan]

hi @angpt ,

thanks for creating this discussion and sorry for the late reply, please feel free to create an pr in camel to integrate any-llm, we would review and add this feature once pr is ready

[aniruddhaadak80]

From my point of view, this seems worth merging if the backend contract stays narrow and capability discovery is explicit.

The thing I would watch is whether any-llm preserves enough surface area for streaming, structured outputs, tool calling, and role semantics without forcing CAMEL into the lowest common denominator. A unified backend is great until subtle provider differences start leaking into agent behavior in hard to debug ways.

If the wrapper can declare supported capabilities clearly rather than pretending every provider behaves the same, I think this could be a clean integration.