Introduction to the Agent Hub
The Agent Hub is the central command center for training your AI agent. This is where you will define the two core aspects of your agent's operation: how it communicates and what it knows.
This guide will walk you through the Agent Hub interface, showing you how to set up communication channels (like SMS, Instagram, and Email) and how to train your agent's knowledge base so it can handle conversations effectively. We will cover both the Knowledge Base for structured data and the interactive Fine-Tuning chat for conversational training.
Accessing the Agent Hub
To begin training a specific AI agent, you first need to navigate to its Agent Hub.
Step-by-Step Instructions:
From the main dashboard, click on AI Agents in the left-hand navigation menu.
This will display all of your created agents. Locate the agent you wish to train.
On that agent's card, click the Agent Hub button.
You will now be taken to the Agent Hub canvas for that specific agent.
The Agent Hub Canvas: Setting Communication Channels
The Agent Hub opens to a visual canvas where you connect your agent to various communication platforms.
Key Concepts:
Agent Trigger: This is the starting point. When this trigger is activated, it initiates communication across all connected channels.
App Dock: The row of icons at the bottom of the screen represents all available communication channels (iMessage & SMS, WhatsApp, Instagram, Messenger, Email, etc.).
Multi-Channel Communication: The agent will send messages through every single app connected to the Agent Trigger on the canvas.
How to Add and Configure a Channel:
Click the icon for the desired app in the App Dock (e.g., click the green iMessage icon for SMS).
The app will appear on the canvas as a "node," automatically linked to the Agent Trigger.
Configure the app node by selecting a Provider from the dropdown menu (e.g., GoHighLevel or Twilio).
For SMS (GoHighLevel): The "Sending Number" field is optional.
Leave it blank to use the default phone number set in your GoHighLevel account.
To use a different GoHighLevel number for this agent, copy and paste that specific phone number into this field.
You can add multiple apps like Instagram, Messenger, and Email to the canvas, and the agent will communicate across all of them when triggered.
Training Your AI's "Brain"
Once you are inside the Agent Hub, you can access the agent's training center, which is its "brain" or memory.
Instructions:
In the top-left corner of the Agent Hub, click the Train AI button.
A pop-up window will appear with two tabs: Knowledge Base and Fine-Tuning.
1. Knowledge Base Tab
This is where you provide your agent with structured, factual information that it can refer to during conversations.
Purpose: To build a foundational understanding of the business.
Sections Include:
Business Information (hours, location, etc.)
Products & Services
Competitor Analysis
Frequently Asked Questions
Objection Handling
Testimonials
Customer Personas
Simply click into any section and fill out the relevant details to build the agent's core knowledge.
2. Fine-Tuning Tab
This is an interactive chat interface where you can test your agent's knowledge and teach it new things conversationally.
Purpose: To test the agent, add one-off pieces of information, and refine its responses.
βHow to Use Fine-Tuning:
Click the Fine-Tuning tab.
Test the Agent: Type a question into the message box to see what it already knows.
Teach the Agent: If the agent doesn't know the answer, provide the correct information. The AI will understand your response and ask for confirmation to save it.
Save to Memory: Reply with "yes" to add the new information to the agent's memory for future conversations.
β
Agent's Memory Panel:
On the right side of the Fine-Tuning chat, you can see the "Agent's Memory." When you ask a question, this panel shows you which pieces of stored information the AI found most relevant to formulate its answer, each with a relevance score.
You can manage the memory directly from this panel by clicking the trash can icon to delete any outdated or incorrect information.
