Chat Agent Smart Component

The Chat Agent Smart Component adds a conversational AI interface to an application. End-users type questions and receive responses from an AI agent configured through the Build Agent. The Chat Agent can optionally accept file attachments, fill form fields on the end-user's behalf, download transcripts, and log conversations to a database. Responses stream to the end-user using HTTP Server-Sent Events.

Creators configure the Chat Agent through the Build Agent, not by editing component properties by hand in the Module Builder.

Before You Begin

Before adding a Chat Agent to a module, confirm the following are in place:

• Access to the Build Agent in the environment.

• At least one LLM credential registered in the environment. The Build Agent lists available credentials during setup.

Create a Chat Agent Smart Component

To add a Chat Agent Smart Component to a module:

1. Open the module in the Module Builder.

2. Open the Build Agent.

3. Describe the chat experience you want to build. Include what questions the agent should answer and any capabilities you want, like file attachments or form fill.

   

4. The Build Agent asks which model the Chat Agent should run on. Available options are determined by the LLM credential configured in the environment.

5. The Build Agent asks which capabilities to enable for the Chat Agent.

6. The Build Agent creates an LLM configuration and adds the Chat Agent Smart Component to the canvas.

Note: The Build Agent asks about the following capabilities during setup: file attachments, transcript download, thinking display, form fill, and conversation logging. You can change these later through the component's settings panel.

For more on working with the Build Agent, see Build Agent Reference.

Configuration

The Configuration tab in the settings panel exposes the settings below. The Build Agent configures these during initial setup.

How to read settings tables

Values are formatted as follows: inline code — a value to type exactly as shown | Bold — a value selected from a drop-down, radio button, tab, or choice chip | ON/OFF — toggle state

Identity and Welcome

The Build Agent sets these fields during setup based on the chat experience description. They can be updated in the settings panel at any time.

LLM Connection

Note: The LLM Config holds the agent's system prompt and model settings server-side. End-users cannot access or change it. To update the prompt, model, or response tuning, ask the Build Agent to revise the configuration.

Tip: If the Chat Agent produces no response, the most common cause is a missing or unconfigured LLM credential. Confirm with the environment administrator that a credential is registered in the environment before troubleshooting the component.

AI Model Settings

The Chat Agent's model, system prompt, and response behavior are defined in an LLM Config managed by the environment administrator in LLM Configs Administration. The Build Agent creates this config automatically during Chat Agent setup. To change the agent's persona, model, or response tuning, ask the Build Agent to update the configuration. Alternatively, work with the environment administrator to edit the config in LLM Configs Administration.

The LLM credential that authorizes requests to the AI provider is registered separately in LLM Credentials Administration. At least one credential must be registered in the environment before a Chat Agent can respond. Supported vendors: Google Vertex AI, Amazon Bedrock, Microsoft Azure OpenAI, and Anthropic.

Each Chat Agent instance is connected to a single LLM credential and model. It cannot switch vendors or models during a conversation. Multiple Chat Agents in the same application can each use a different LLM config, credential, and vendor.

Capabilities

The settings panel displays only the capabilities selected during Build Agent setup. All capabilities can be toggled on or off at any time from the settings panel, regardless of what was chosen during setup.

Note: When Enable Form Writes is ON, the agent can set component values and attach files to components in the module. The agent cannot submit the form. End-users review the populated values and submit as usual.

Appearance

The Appearance tab controls the color scheme for the Chat Agent.

Note: The Chat Agent's position in the module is determined by the Build Agent during setup, not through a component setting. The Build Agent can place the Chat Agent as a drawer at the edge of the viewport or inline within the form. To change placement, ask the Build Agent to update the configuration.

Data

The Data tab controls how the Chat Agent stores data in the module.

Note: Store Data in Database saves the component's field value as part of the module submission. This is separate from Enable Conversation Logging, which saves full conversation transcripts as submissions in a target module.

File Attachment Limits

When Enable File Attachments is ON, end-users can attach the following file types to their messages.

Files that exceed these limits are rejected before they reach the agent.

Note: The agent reads file content during the session. File bytes are not stored by the platform after the session ends.

Data Access

The Chat Agent automatically reads component data from the module in which it is placed. No configuration is required. During an end-user session, the agent can read the current value, label, and state of every component in the module.

The agent does not have access to the following by default:

• Past submissions from the same end-user or other end-users.

• Data from modules outside the current screen.

• The module's configuration or structure.

The agent cannot make direct requests to external systems. To connect an external API, build a Unqork module that proxies the external service and instruct the Build Agent to use that module when answering the relevant questions. The agent selects which API module to call based on user intent and the instructions in its system prompt.

Note: If a use case requires the agent to answer questions about submitted records, data from a previous session, or data from another module, that data must be made available through API modules built by the application team.

Note: Creators are responsible for verifying that any backend data access module generated by the Build Agent implements the correct filtering, access controls, and input validation. The Build Agent generates the module structure, but data security must be confirmed through standard testing before deploying to end-users.

Conversation Logging and Privacy

Warning: When Enable Conversation Logging is ON, every end-user message and every agent response is stored as a submission. This includes the full text of each conversation. Evaluate your privacy requirements and applicable data regulations before enabling this setting.

Conversations are saved as submissions in the module set by Audit Submissions Module ID. When that field is blank, conversations log to the module where the Chat Agent renders.

Only the user-visible turns are stored. Reasoning traces and internal tool activity do not appear in the log.