Currently, the AI Toolkit allows for tool usage if the provider exposes that functionality. However, it's a big pain to manage the history in that case, since the tool calls and results have no mappings and you want to keep those in the history.

And, the tool handling changes based on what LLM you're using.

But worst of all, if the LlmChatView or app in some other way calls DartanticProvider.history = ..., then all of the tool calls in the history will be lost.😭