Using Agent

Chat Interface

How to use the chat interface to interact with your agent

What is Chat Interface?

Chat Interface

Chat Interface for Testing Agent

Chat Interface is the page where you can talk directly with your agent for testing and development. This interface provides the same chat experience that your end users will have.

Main Uses:

  • Testing agent before publishing
  • Try different questions and scenarios
  • See agent response and behavior
  • Debug actions and knowledge base
1

How to Access Chat

Open your agent's Overview page, then click the "Test Agent" button or chat icon at the top.

1
From Dashboard

Click agent card → Overview → Test Agent

2
From Agent List

Click the chat icon on agent card

2

Chat Interface Features

Message Input

Type questions or instructions for the agent. Press Enter or click send button.

Message History

All conversations are saved automatically. Scroll up to see previous chats.

Agent Response

Agent will reply based on knowledge base and conversation context.

Actions Execution

If agent needs to call action, you will see loading indicator.

🔐 Persistent Context

If your agent has Persistent Context enabled, the chat experience will be different - the agent will "remember" who the user is and maintain conversation context even across different sessions.

Automatic Login

When a user first chats (or after session expires), the agent will automatically request login if there are Private Actions requiring context:

User: Check my balance

Agent: To view your balance, I need you to log in first. Please enter your email and password.

Persistent Session

After logging in, user context is saved and available for all subsequent conversations:

User doesn't need to log in again for each chat
Private Actions can be executed immediately
Agent uses user data for personalization
Session Management: Fully Persistent (default) or Time-Based Expiration

Two Session Modes:

  • Fully Persistent: Session never expires, active until manual logout
  • Time-Based: Token expires after set time, can auto-refresh
Private Actions

Actions requiring user context can only be executed after user logs in:

User: Transfer 100k to 08123456789

Agent: Alright, I will transfer 100,000 from your account (****1234) to 08123456789...

Transfer successful! Agent used user_id from context.

Logout

Users can log out anytime, which will clear context and require re-login:

User: Logout

Agent: You have been logged out successfully. See you again!

Persistent Context Testing Tips
  • Test login flow with both valid and invalid credentials
  • Try Private Actions before and after logging in
  • Test logout and re-login to verify session management
  • Use incognito browser to simulate new users
  • Verify that context data is used correctly in Private Actions
3

Understanding Agent Response

Knowledge Base Response

Agent answers based on information from knowledge base you added.

User: What are your operating hours?

Agent: We are open Monday-Friday 09:00-17:00 WIB.

Response with Action

Agent executes action and provides feedback on execution result.

User: Please create a support ticket

Agent: Ticket created successfully with ID #12345

No Information Response

If agent doesn't find relevant information, it will inform the user.

User: What is the price of product X?

Agent: Sorry, I don't have information about that product.

Agent Testing Tips

Test Various Scenarios

Try easy, difficult, and edge case questions to ensure agent handles well

Check Context Awareness

Test if agent remembers previous conversation in the same session

Verify Actions

Ensure actions are called with correct parameters

Verify Knowledge Base

Ensure all knowledge bases are trained with Ready status

4

Managing Chat

Start New Conversation

Click "New Chat" button to start fresh conversation without previous context.

Delete Chat

Click delete icon on chat to remove unnecessary conversations.

Chat History

All chats are saved and can be accessed anytime from sidebar.

Troubleshooting

Agent Not Responding

Ensure agent has trained knowledge base (status Ready)

Inaccurate Answers

Check if knowledge base is complete and up-to-date

Action Failed

Check action configuration (URL, headers, authentication) and ensure API endpoint works