If using `MultiLLMPromptExecutor` what determines ...

# koog-agentic-framework

[John O'Reilly]

If using MultiLLMPromptExecutor what determines when each one gets used? Also, which models get used in this case?

val multiExecutor = MultiLLMPromptExecutor(
        LLMProvider.OpenAI to openAIClient,
        LLMProvider.Google to googleClient,
    )

[Didier Villevalois]

The model used is the one you give to your AIAgentConfig when creating your AIAgent. When creating a sub-graph, you can override that model for the sub-graph. (You can also manually override the current model inside a custom node, by setting it in an llm.writeSession { } block.)

The MultiLLMPromptExecutor only serves as a "control tower" to identify the correct LLMClient for the used model. (i.e. an LLModel has a provider, the MultiLLMPromptExecutor can resolve a client for this provider.)