Overview
Real-time messaging
Instant responses with streaming support
AI agent integration
Connect to custom agents or FlowX accelerators
Session management
Persistent conversations across sessions
Knowledge base access
Context-aware responses from your data
How it works
The Chat component follows a simple flow:User initiates chat
User opens your application and the Chat component loads with a welcome message.
Agent processing
The agent processes the message, optionally querying knowledge bases or external systems.
Key concepts
Agent workflows
The Chat component connects to an agent workflow that defines how messages are processed:- Input handling - How user messages are received
- AI processing - Which models and prompts are used
- Knowledge access - What data sources the agent can query
- Response generation - How responses are formatted
Session persistence
Conversations can be persisted to maintain context:| Mode | Description | Use case |
|---|---|---|
| Session-based | Conversation lasts for browser session | Quick queries, support |
| Persistent | Conversation saved to database | Ongoing cases, history |
| Process-linked | Tied to a specific process instance | Guided workflows |
Knowledge base integration
Agents can access knowledge bases for informed responses:- RAG (Retrieval-Augmented Generation) - Agent searches relevant documents before responding
- Semantic matching - Finds conceptually similar content, not just keyword matches
- Source attribution - Responses can cite their sources
Knowledge base integration requires configuration in the agent workflow. See Knowledge Base Overview for setup details.
Display modes
The Chat component supports different display modes:| Mode | Description |
|---|---|
| Standalone | Full-page chat experience |
| Embedded | Chat panel within a larger UI |
| Floating | Expandable chat button overlay |
Best practices
Set clear expectations
Set clear expectations
Use the welcome message to explain what the agent can help with. This reduces user frustration from mismatched expectations.
Provide context
Provide context
Pass relevant process data to the agent (user info, current step, etc.) so responses are contextually appropriate.
Handle limitations gracefully
Handle limitations gracefully
Configure fallback responses for when the agent can’t help. Offer escalation to human support when needed.
Design for mobile
Design for mobile
Test the chat experience on mobile devices. Ensure the interface works well with on-screen keyboards.
Implementation
For detailed configuration and styling options, see the UI Designer documentation:Chat component configuration
Complete configuration reference for the Chat UI component