Skip to main content

AI Generate Reply

The AI Generate Reply node uses your configured AI provider to generate text. Give it instructions and context, and it writes a reply for you. This is the core node for building AI-powered conversations.

AI Generate Reply

Configuration​

System Prompt​

This sets the AI's personality and ground rules. Think of it as the instructions you'd give a new employee on their first day.

Example:

You are a helpful assistant for ABC Clinic. Keep responses short and friendly. Always include the clinic phone number 555-1234.

User Prompt​

This is the actual request you're sending to the AI. It supports {{variables}}, so you can inject contact data, message content, and chat history.

Example:

Customer {{name}} sent: {{incoming_message}}

Previous conversation: {{chat_history}}

Write a helpful reply.

Max Tokens​

Limits how long the AI's response can be. The default is around 500 tokens (roughly a few paragraphs). Lower this if you want shorter replies.

Temperature​

Controls how creative or predictable the AI is:

  • 0 — Deterministic and consistent. The AI gives the same answer every time for the same input.
  • 0.7+ — More creative and varied. Good for casual conversations where you want natural-sounding variety.

Model​

By default, this node uses whichever model you've set in Settings. You can override it here if you want a specific node to use a different model (for example, a faster model for simple tasks or a more capable one for complex replies).

Model Settings

Output​

The generated text is available as {{ai_reply}} for any downstream nodes. You'll typically connect this to a Send Reply or Send Message node.

Requirements​

You need an AI provider configured in Settings before this node will work. See the AI Setup page for setup instructions.

Error Handling​

If the AI call fails (network issues, rate limits, etc.), the node automatically retries up to 3 times with exponential backoff. This means it waits a little longer between each retry, giving the API time to recover.

Tips​

  • Keep your system prompt focused. Tell the AI what it is, what it should do, and what it should never do.
  • Use {{chat_history}} in your user prompt so the AI has context about the conversation. Pair this with the Get Chat Messages node upstream.
  • Start with temperature 0 while testing, then increase it once you're happy with the results.