Workflow examples for google gemini models

[networkandcode]

Are there any multi agent example with gemini models.. I see the examples contain openai references, just wondering if that work with gemini as it has open ai compatible API?

[saqadri]

@networkandcode yes they should all work with gemini as well! You can just replace OpenAIAugmentedLLM references with GoogleAugmentedLLM and configure your google gemini API keys and the same examples will just work. Here's an example how to configure google gemini settings:

https://github.com/lastmile-ai/mcp-agent/blob/main/examples/model_providers/mcp_basic_google_agent/mcp_agent.config.yaml

Let me know if you need any help!

[networkandcode]

@saqadri I tried this already, I was checking if workflow eample would work with this for ex. OpenAI router... let me give a try, thx.

[saqadri]

@networkandcode I see what you mean! The Router workflow has two specific helpers for Anthropic and OpenAI, but they are purely syntactic sugar. You can very easily use Router workflow with Google Gemini this way:

return LLMRouter.create(
    name=name,
    llm_factory=GoogleAugmentedLLM, <-- this is key
    server_names=server_names, # if any
    agents=agents, # if any
    functions=functions, # if any
    routing_instruction=routing_instruction, # your prompt
    context=context,
    **kwargs,
)

BTW there's also https://github.com/lastmile-ai/mcp-agent/blob/main/src/mcp_agent/workflows/factory.py#L119 which lets you construct things more easily. I also just realized thanks to your question that there's a bug in create_router_llm where it will error if a non-openai/anthropic model is provided. My above code will get you unblocked, and I'll fix create_router_llm separately.