Observations in SmartFlow
Observations provide real-time feedback to your AI assistant during conversations. They allow SmartFlows to communicate results back to the AI.
What Are Observations?
When a SmartFlow executes (triggered by a Conversation Tool), it can send observations back to the AI. The AI uses these observations to:
- Provide informed responses to users
- Display data retrieved from external systems
- Confirm actions were completed successfully
- Handle errors gracefully
How It Works
User asks question → AI triggers SmartFlow → SmartFlow executes →
Observation sent → AI reads observation → AI responds to user
Example:
- User: "What's the status of my order #12345?"
- AI triggers the "Check Order Status" SmartFlow
- SmartFlow calls the eCommerce API
- SmartFlow sends observation:
{"status": "shipped", "tracking": "ABC123"} - AI responds: "Your order #12345 has shipped! Tracking: ABC123"
Setting Observations
Use the Set Observation action node in your SmartFlow:
| Field | Description |
|---|---|
| Observation | Text or JSON data to send to the AI |
| Variables | Use {{variable}} to include dynamic data |
Example configuration:
{
"status": "{{apiResponse.orderStatus}}",
"tracking": "{{apiResponse.trackingNumber}}",
"estimatedDelivery": "{{apiResponse.eta}}"
}
Best Practices
Keep Observations Concise
The AI processes observations as context. Keep them focused:
// Good
{"found": true, "productName": "Blue Widget", "price": 29.99}
// Too verbose
{"query": "...", "timestamp": "...", "requestId": "...", "fullResponse": {...}}
Handle Errors
Always include error handling in your observations:
// Success
{"success": true, "data": {...}}
// Error
{"success": false, "error": "Order not found"}
Use Structured Data
JSON is preferred over plain text for complex data:
{
"tickets": [
{"id": "JIRA-123", "status": "In Progress"},
{"id": "JIRA-124", "status": "Done"}
],
"total": 2
}
Common Use Cases
| Use Case | Observation Contains |
|---|---|
| Order lookup | Status, tracking, delivery date |
| Appointment booking | Confirmation, date/time, details |
| Knowledge search | Relevant articles, summaries |
| CRM lookup | Customer info, history |
| Inventory check | Stock levels, availability |
Related
- Set Observation Action — Node reference
- Conversation Tool Trigger — How AI triggers flows
- AI Request Action — Request AI processing