Add a Chat Agent to an Interface

Adding a chat agent to an Appian interface lets end users interact with the chat agent through a real-time conversational experience. The a!agentChatField() component handles messaging, streaming responses, and conversation persistence without requiring manual orchestration.

Prerequisites for adding a chat agent to an interface

Adding a chat agent to an interface requires the following:

• Chat agent: A chat agent must be created and configured in Agent Studio. The agent type is selected at creation and cannot be changed later.
• Tool permissions: The user interacting with the chat agent must have permissions to all tools the agent has access to. A chat agent can also operate without any tools configured.
• Interface object: An existing or new interface where the agent chat component will be added.

Add the agent chat component to an interface

Adding the agent chat component to an interface connects end users with a chat agent for real-time conversational interactions. End users can ask questions, receive streamed responses, and interact with the chat agent's capabilities directly within the application.

To add the agent chat component to an interface:

1. In Appian Designer, open or create an interface object.
2. In expression mode, type a!agentChatField to add the component. You can also drag the agent chat component from the component palette.
3. In the agent parameter, reference your chat agent object. For example: agent: agent!MY_CHAT_AGENT.
4. (Optional) If you need to provide context-specific data to the chat agent, configure the inputs parameter. Pass a map of key-value pairs that correspond to the inputs defined in Agent Studio. For example:

   1
   2
   3
   inputs: {
     caseId: ri!caseId
   }
   

5. (Optional) If you need to capture data produced by the chat agent, configure the outputsSaveInto parameter. Use save!value to reference the output variables. For example:

   1
   outputsSaveInto: a!save(local!agentDecision, save!value.decision)
   

6. (Optional) If you need to see tool inputs and outputs for troubleshooting during development, set debugMode: true. You can also view agent runs in the AI agent design object for additional information. Disable debug mode before deploying to production.
7. Click SAVE to save the interface.

The agent chat component renders in the interface and is ready for testing.

Configure the agent chat component appearance

The agent chat component supports standard parameters for controlling its visual appearance within the interface layout.

To configure the agent chat component appearance:

1. Set the height parameter to control the vertical space. The default value is "FILL", which makes the component expand to fill the available vertical space. Other valid values include "EXTRA_SHORT", "SHORT", "SHORT_PLUS", "MEDIUM", "MEDIUM_PLUS", "TALL", "TALL_PLUS", and "EXTRA_TALL".
2. (Optional) If you want placeholder text in the input area, set the placeholder parameter.
3. (Optional) If you want to display introductory text before the user sends their first message, set the welcomeMessage parameter.

Test the agent chat component

Testing the agent chat component verifies that a!agentChatField() renders correctly and the chat agent responds as expected.

To test the agent chat component:

1. In Appian Designer, open the interface containing the agent chat component.
2. Click Preview to render the interface.
3. Type a question in the input area and press Enter to send the message.
4. Verify that the chat agent responds with streamed output and displays its tool calls.
5. (Optional) If debug mode is enabled, verify that tool inputs and outputs display alongside responses.
6. (Optional) Review agent runs in Agent Studio and provide developer feedback on the run.

Agent chat component use cases

The agent chat component supports a variety of conversational patterns. The following scenarios illustrate common implementations:

• Form-filling guidance: Embed the agent chat component alongside a complex form to provide real-time guidance on technical fields. Pass the current form data as inputs so the chat agent can validate entries and suggest corrections.
• Case history summarization: Connect a chat agent to case record data. End users can ask the chat agent to summarize information from multiple sources (emails, records, documents) without manually reviewing each one.
• Multi-source search: Configure a chat agent with tools that search multiple data sources (documents, records, databases). End users ask one question and receive a comprehensive answer aggregated from all available sources.

For the full component reference, including the parameter table and additional examples, see Agent Chat Component.

Feedback