Dmytro Skarzhynets
1 year ago
4 changed files with 86 additions and 7 deletions
@ -1 +1,77 @@ |
|||||
### Prompt settings |
#### Example Usage: AI-Powered Alarm Analysis |
||||
|
|
||||
|
This example demonstrates how to use the AI node to automatically analyze a new device alarm, generate a human-readable summary, and suggest troubleshooting steps. |
||||
|
This is useful for creating enriched notifications or populating a dashboard widget. |
||||
|
|
||||
|
#### Scenario |
||||
|
|
||||
|
An IoT freezer unit generates a "High Temperature" alarm. We want the AI to process this alarm data to create a clear summary and a recommended action plan for an operator. |
||||
|
|
||||
|
1. Incoming Message Structure |
||||
|
|
||||
|
When the alarm is created, the message sent through the rule chain has the following structure: |
||||
|
|
||||
|
Message body (represent an alarm, usual alarm fields omitted for brevity) |
||||
|
```json |
||||
|
{ |
||||
|
"details": { |
||||
|
"currentTemp_C": -5, |
||||
|
"threshold_C": -18 |
||||
|
} |
||||
|
} |
||||
|
``` |
||||
|
|
||||
|
Message metadata |
||||
|
```json |
||||
|
{ |
||||
|
"deviceName": "Freezer-B7", |
||||
|
"deviceType": "CommercialFreezer" |
||||
|
} |
||||
|
``` |
||||
|
|
||||
|
2. To achieve our goal, we configure the two prompt fields as follows: |
||||
|
|
||||
|
**System prompt** |
||||
|
|
||||
|
Here, we set the AI's role and enforce a strict JSON output format. This ensures the output is always machine-parsable. |
||||
|
|
||||
|
``` |
||||
|
You are an expert AI assistant for IoT operations. |
||||
|
Your task is to analyze device data and respond with a single, valid JSON object. |
||||
|
Do not include any text, explanations, or markdown formatting before or after the JSON output. |
||||
|
``` |
||||
|
|
||||
|
**User prompt** |
||||
|
|
||||
|
This prompt defines the specific task, using templates to dynamically insert data from the incoming alarm message. |
||||
|
|
||||
|
``` |
||||
|
Analyze the following alarm from a "${deviceType}" unit named "${deviceName}". |
||||
|
|
||||
|
Alarm Data: |
||||
|
$[*] |
||||
|
|
||||
|
Based on the alarm data, generate a JSON object with two keys: |
||||
|
1. "summary": A brief, human-readable summary of the event. |
||||
|
2. "action": A concrete, recommended next step for an operator. |
||||
|
``` |
||||
|
|
||||
|
3. How It Works |
||||
|
|
||||
|
When the alarm message from "Freezer-B7" is processed by the AI node, the templates are substituted with the actual data: |
||||
|
- `${deviceName}` becomes "Freezer-B7" |
||||
|
- `${deviceType}` becomes "CommercialFreezer" |
||||
|
- `$[*]` is replaced by the entire message body JSON: `{"alarmType": "High Temperature", "severity": "CRITICAL", "currentTemp_C": -5, "threshold_C": -18}` |
||||
|
|
||||
|
The final instruction sent to the AI is a combination of the system and the substituted user prompt. |
||||
|
|
||||
|
4. Expected AI Output |
||||
|
|
||||
|
Given the combined instructions, the AI would generate the following structured JSON output, which can then be used in subsequent rule nodes (e.g., to send an enriched email or create a trouble ticket). |
||||
|
|
||||
|
```json |
||||
|
{ |
||||
|
"summary": "Critical high temperature alert on freezer unit Freezer-B7. The current temperature is -5°C, which is significantly above the required threshold of -18°C.", |
||||
|
"action": "Dispatch technician immediately to inspect the unit's cooling system and ensure the door is properly sealed. Investigate for potential power issues." |
||||
|
} |
||||
|
``` |
||||
|
|||||
Loading…
Reference in new issue