Skip to main content

​
AI Agents in LyznFlow

Building an AI Agent is one of the most powerful features of LyznFlow. It transforms a standard workflow into an intelligent conversational bot.

​
The Agent Architecture

An AI Agent workflow consists of three key components working together:

​
1. Chat Trigger

This is the entry point. It listens for messages from the chat widget.
  • Session ID: Automatically generated for each user conversation.
  • Message: The user’s text input.
  • History: Previous messages in the conversation (handled automatically).

​
2. AI Agent Node

The brain of the operation.
  • System Prompt: You define the persona (e.g., “You are a helpful customer support assistant”).
  • Model: Choose the LLM (e.g., GPT-4, Claude).
  • Memory: The agent remembers the conversation context using the Session ID.

​
3. Tools

Tools give your agent “hands”. Without them, it can only talk. With them, it can:
  • Search the web (Live Search).
  • Check your database (LyznDB Query).
  • Send emails (Send Email Tool).

​
Building Your First Agent

  1. Start Needs: Drag a Chat Trigger onto the canvas.
  2. Add Brain: Connect it to an AI Agent node.
  3. Equip Tools:
    • Add tool nodes (e.g., Google Search Tool).
    • Connect the tool’s output to the AI Agent’s “Tools” input port.
    • You can connect multiple tools to the same Agent!

​
The Execution Loop

When a user asks a question like “What’s the weather in Tokyo?”:
  1. AI Thinks: The Agent receives the message.
  2. Tool Decision: It realizes it doesn’t know the weather, but it has a Weather Tool.
  3. Tool Call: It pauses execution and calls the tool with { location: "Tokyo" }.
  4. Tool Result: The tool runs and returns "Sunny, 25°C".
  5. Final Response: The AI uses this data to answer: “It is currently sunny and 25°C in Tokyo.”
Next Steps: Want to build your own custom tools? Check out the Custom Tool Builder.