Skip to main content
The language model reads the conversation so far and decides what the agent says next. It also evaluates your edge conditions, so it drives when the conversation moves from one node to the next. Set your default model on Model Configurations, and override it per workflow where you need something different. Credentials come from a connection.

​
Supported providers

A provider listed here as covering more than one service needs only one connection. Connect OpenAI once and it can drive the language model, the voice, and the transcriber.

​
Choosing a model

Each provider ships a list of models to pick from. If the one you want is not listed, enter its model name manually — the field accepts any model your account can reach. Faster models cost less and reply sooner, which matters more in a voice conversation than in chat. A slow model is heard as a pause. Start with a fast model and move up only if the agent is not keeping up with your prompt.