Choose AI and reply behavior

Choose an AI mode

• Doublewire model uses a provider connection managed by Doublewire. You do not enter a provider key.
• Bring your own key lets you choose an available provider and model and enter that provider’s API key.

Available models can change. When you switch a provider, review the model selection again. Provider pricing shown in the form describes the provider’s token charge for workspace-provided keys.

Control reply behavior

• Temperature controls permitted response variation. Some models do not support it.
• Max Tokens limits the length of one response.
• Memory controls how much recent conversation history is included in a reply request.
• Max Conversation Length limits the total number of messages in one conversation before the visitor must start a new chat.

Memory and conversation length solve different problems: one controls current AI context, while the other controls the lifetime of the conversation.

Use contact memory in replies

Enable this setting to include available contact memory in the assistant’s prompt. Collection must also be enabled on the channel, and the workspace must have access to the feature.

Always test meaningful scenarios in Preview after changing model, prompt, or advanced behavior.