Skip to main content
All CollectionsFlows
Understanding the "Flow" designer
Understanding the "Flow" designer

Empower your AI Assistant to take action and get work done with a flexible and simple platform for agentic workflows.

Alvaro Vargas avatar
Written by Alvaro Vargas
Updated over 2 months ago

The designer is where users can build powerful flows that integrate their AI Agent with key business systems. Important: Everything that happens within the designer like dropping a new node, updating it's content, etc., will be automatically saved using auto-save.

The designer has several key components:

The canvas

This is the place where actions or nodes are placed and connected. The Canvas has controls, like: zoom in, zoom out, mini map, and lock (prevents you from moving nodes across the canvas).

Actions or nodes

Actions or nodes are steps in a sequence, there are several types of actions (and this number will be growing quickly). Each step is executed sequentially.

Node Types

  1. Intent: the intent action is typically used as the trigger for a flow (E.g.: "Where is my order").

  2. AI Assistant: the AI Assistant node allows you to engage the user with an AI Assistant that contains specific configs like: custom instructions, model, temperature, and variables to capture during the conversation (e.g.: an order number). The advantage of using the AI Assistant node is that using the "Where is my order" example, the AI Assistant can not only request an order number but also help the user find their order number and provide guidance and valuable information through instructions and it's knowledge base.

  3. Say: this node allows you to output a single message. You can output a hardcoded message or prompt an AI model to output a message. For example if you are expecting a value like "Delivered" to stored within a variable, you can use a hardcoded message, e.g.: "Your order status is: {order_status}". If instead you want to store an entire output from an API call in a variable, like all the properties of a lead and you want the AI to interpret the API response and output a message based on this, using a prompt, then Say with AI is the best option. E.g.: "Let me know if this lead is qualified or not based on our lead scoring system {lead_score_system} {lead}"

  4. API Call: the API call node allows you to connect your agent to any external system like a CRM, Ticketing System, ERP and more, via REST APIs. Our approach ensures reliability and security since the API call is made by our backend and not by the AI Agent using function calling.

Adding a node to the canvas

Adding a node to the canvas is very simple, just drag and drop the node from. the left menu and into the canvas.

Deleting a node from the canvas

There are three ways to delete a node:

  1. Clicking on the node and pressing the "delete" key.

  2. Clicking on the three dots on the top right corner of the node, and selecting "delete" from the dropdown menu.

  3. Opening the node's details, and selecting the 3 dots on the top right corner of the modal. Select "delete" from the dropdown menu.

Connecting nodes

Nodes have an input connector (left side of the node) and an output connector (right side of the node). Nodes can be connected by clicking (and holding down) on the output connector and dragging the connecting line to and input connector of another node. Note that for now, nodes can connect to a single node at a time. If you attempt to connect a node to multiple nodes, e.g.: a AI Assistant node to an API Call node and a Say node, the last connection will be the one executed.

Deleting a connection between nodes

To delete connections between node, click on the connecting line between nodes and press the delete key on your keyboard.

Header

Flow Status

Flows have 3 states: Live, Offline and Draft.

If a Flow is set to "Live", the AI Assistant will be listening for triggers and execute the flow whenever the trigger is fired. Your flow will be live in production.

If a Flow is set to Offline, the AI Assistant won't be listening for triggers and the flow won't be executed. Important: If set to Offline, the flow won't be executed even within the Frontline dashboard for testing purposes.

If a Flow is set to Draft, the AI Assistant will only listen for triggers within the Frontline dashboard, which makes it perfect for testing.

Testing your AI Assistant

To test your AI Assistant from the designer, click on the "Test" button on the top right corner. Once the Live Chat interface is loaded, you can interact with your AI Assistant. Note that the interaction won't be just contained to the specific flow you are viewing, but you'll be able to test the entire assistant. To test a specific flow, you can trigger the flow by sending a message that matches the intent.

Navigating to a different flow

To navigate to a different flow without leaving the designer, click on the dropdown menu with the flow's name (top center), and select the flow that you want to navigate too.

Did this answer your question?