跳到主要内容
版本：v1.7.x

AI Agent Chat User Guide

HertzBeat AI Agent Chat is an intelligent monitoring AI Agent that helps you manage monitors, configure alerts, and optimize your infrastructure monitoring through natural language conversation.

Overview​

The AI Agent Chat feature provides an interactive chat interface where you can:

Monitor Management:

  • 🔍 Query and filter existing monitors by status, type, host, and labels
  • ➕ Add new monitors for websites, APIs, databases, and services
  • 📊 Get detailed information about available monitor types and their parameters
  • ⚡ Check monitor status and troubleshoot monitoring issues

Alert Management:

  • 🚨 Query active alerts with comprehensive filtering (type, status, search)
  • 📈 Get alert summary statistics and distribution
  • 🔔 View both single and grouped alerts
  • 📋 Analyze alert patterns and trends

Metrics Analysis:

  • 📊 Retrieve real-time metrics data for any monitor
  • 📈 Access historical metrics with customizable time ranges
  • 💾 Check warehouse storage system status
  • 🔍 Query specific metric fields and labels

Alert Rule Management:

  • ⚙️ Configure alert rules and thresholds
  • 📝 Manage alert definitions for different monitor types
  • 🎯 Set up custom alerting criteria

Prerequisites​

Before using the AI Agent Chat, ensure:

1 OpenAI Configuration: Valid OpenAI API key must be configured. Optionally, you can change the model in application.yml under spring.ai.openai.chat.options.model (default is gpt-4.1-nano). 2 Database Connection: HertzBeat database must be accessible for monitor operations

Configuration​

OpenAI API Key Setup​

The AI Agent Chat uses OpenAI's GPT models. You need to configure an OpenAI API key in one of two ways:

  1. Navigate to the AI Agent Chat interface
  2. If no API key is configured, you'll see a configuration dialog
  3. Enter your OpenAI API key
  4. Click Save Configuration

Method 2: YAML Configuration​

Add to your application.yml:

spring:
  ai:
    openai:
      api-key: sk-your-openai-api-key-here

Configuration Priority​

  • Database configuration takes priority over YAML configuration
  • Configuration changes are applied immediately without restart
  • Invalid API keys are automatically validated before saving

Using the AI Agent Chat​

Accessing the Chat Interface​

  1. Navigate to the AI Agent Chat section in HertzBeat web interface
  2. You'll see the HertzBeat Monitoring AI Agent welcome screen
  3. Start typing your questions in the chat input field

Sample Conversations​

Adding a New Monitor​

You: Add a new HTTP monitor for my website https://example.com
AI Agent: I'll help you add an HTTP monitor for https://example.com. 
Let me create that monitor with the following configuration:
- Name: example.com Website Monitor
- Type: HTTP
- Host: example.com
- Port: 443 (HTTPS)
- Path: /
- Interval: 60 seconds

Monitor created successfully with ID: 12345

Listing Existing Monitors​

You: Show me all my current monitors
AI Agent: Here are your current monitors:

1. **Linux Server** (ID: 1001)
   - Status: ✅ Online
   - Host: 192.168.1.100
   - Last Check: 2 minutes ago

2. **MySQL Database** (ID: 1002) 
   - Status: ⚠️ Warning
   - Host: db.example.com
   - Last Check: 30 seconds ago

3. **Website Monitor** (ID: 1003)
   - Status: ✅ Online
   - Host: https://example.com
   - Last Check: 1 minute ago

Would you like details on any specific monitor?

Getting Monitor Type Information​

You: What parameters do I need to monitor a PostgreSQL database?
AI Agent: To monitor a PostgreSQL database, you'll need these parameters:

**Required Parameters:**
- **Host**: Database server IP/hostname
- **Port**: PostgreSQL port (default: 5432)
- **Database**: Database name to connect to
- **Username**: Database username
- **Password**: Database password

**Optional Parameters:**
- **Timeout**: Connection timeout (default: 6000ms)
- **Query Timeout**: SQL query timeout (default: 3000ms)
- **SSL**: Enable SSL connection (default: false)

Would you like me to help you add a PostgreSQL monitor?

Alert Management​

You: Show me all critical alerts from the last 24 hours
AI Agent: Here are the critical alerts from the last 24 hours:

🚨 **Active Alerts (3 firing)**
1. **High CPU Usage** - Linux Server (ID: 1001)
   - Status: 🔥 Firing
   - Triggered: 2 hours ago
   - Severity: Critical
   - Message: CPU usage 95% > threshold 90%

2. **Database Connection Failed** - MySQL Database (ID: 1002)
   - Status: 🔥 Firing  
   - Triggered: 45 minutes ago
   - Severity: Critical
   - Message: Connection timeout after 10 seconds

3. **Website Response Time** - Website Monitor (ID: 1003)
   - Status: ⚠️ Warning
   - Triggered: 15 minutes ago
   - Severity: Warning
   - Message: Response time 5.2s > threshold 3s

**Alert Summary:**
- Total alerts: 12
- Firing: 3 (2 critical, 1 warning)
- Resolved: 9

Would you like me to help you investigate any specific alert?

Metrics Analysis​

You: Show me the CPU usage for my Linux server over the last 6 hours
AI Agent: Here's the CPU usage data for your Linux server (Monitor ID: 1001) over the last 6 hours:

📊 **CPU Usage Metrics - Last 6 Hours**

Time | Usage (%) | Load Avg | Processes 14:00-15:00 | 45.2% | 1.2 | 156 15:00-16:00 | 67.8% | 2.1 | 178 16:00-17:00 | 89.4% | 3.8 | 195 17:00-18:00 | 92.1% | 4.2 | 203 18:00-19:00 | 94.7% | 4.8 | 210 19:00-20:00 | 88.3% | 3.5 | 188


**Analysis:**
- Peak usage: 94.7% at 19:00
- Average usage: 79.6%
- Trend: High usage sustained for 3+ hours
- Current status: Decreasing (88.3%)

**Real-time metrics:**
- Current CPU: 76.2%
- Memory usage: 82.1%
- Disk I/O: Normal

Would you like me to check historical data for comparison or investigate the high usage cause?

Alert Rule Configuration​

You: Set up an alert rule for when my website response time exceeds 5 seconds
AI Agent: I'll help you create an alert rule for website response time monitoring.

**Alert Rule Configuration:**
- **Monitor Type**: HTTP/Website
- **Metric**: Response Time
- **Condition**: Greater than 5 seconds
- **Severity**: Warning
- **Evaluation**: Check every 60 seconds
- **Notification**: Email + Dashboard

**Proposed Rule:**
```yaml
name: "Website Response Time Alert"
condition: "response_time > 5000ms"
severity: "warning"
description: "Alert when website response time exceeds 5 seconds"
evaluation_interval: 60s
for: 2m  # Alert after condition persists for 2 minutes

This rule will:

  1. Monitor response time every minute
  2. Trigger alert if response time > 5s for 2+ minutes
  3. Send notifications via configured channels
  4. Show in dashboard with warning status

Would you like me to apply this configuration or modify any settings?


### Chat Features

#### Message History

- All conversations are automatically saved
- Access previous conversations from the sidebar
- Search through conversation history
- Export conversation logs

#### Real-time Responses

- Streaming responses for immediate feedback
- Typing indicators show when the assistant is processing
- Cancel ongoing requests if needed

#### Conversation Management

- Create new conversations for different topics
- Rename conversations for better organization
- Delete old conversations to keep things clean
Note: Conversations are not saved in the database

## Troubleshooting

### Common Issues

#### Chat Interface Not Loading

**Symptoms**: Chat interface shows loading spinner indefinitely
**Solutions**:

1. Check browser console for JavaScript errors
2. Ensure network connectivity to HertzBeat server

#### "Service Unavailable" Message

**Symptoms**: Chat shows "HertzBeat AI monitoring service unavailable"
**Solutions**:
1.Verify OpenAI API key configuration
2.Check application logs for errors
3.Ensure database connectivity

#### Invalid API Key Error

**Symptoms**: Configuration dialog shows "Invalid API key" error
**Solutions**:

1. Verify your OpenAI API key starts with `sk-`
2. Check API key has sufficient credits/quota
3. Test API key directly with OpenAI API
4. Ensure no extra spaces in the API key

#### Monitor Creation Failures  

**Symptoms**: AI suggests monitor configuration but creation fails
**Solutions**:

1. Verify you have permissions to create monitors
2. Check if monitor with same name already exists
3. Ensure target host/service is accessible
4. Review monitor parameter validation errors

### Debug Mode

Enable debug logging by setting log level to DEBUG for:

```yaml
logging:
  level:
    org.apache.hertzbeat.ai.agent: DEBUG

Best Practices​

Effective Chat Usage​

  1. Be Specific: "Add HTTP monitor for api.example.com port 8080" vs "add a monitor"
  2. Provide Context: Mention if you want production vs test monitors
  3. Ask Follow-ups: Request configuration details if needed
  4. Use Natural Language: The AI understands conversational requests

Monitor Management​

  1. Naming Convention: Use descriptive monitor names
  2. Documentation: Ask the AI to document complex configurations

Security Considerations​

  1. API Key Security: Store OpenAI API keys securely
  2. Access Control: Limit AI chat access to authorized users
  3. Data Privacy: Be mindful of sensitive information in chat logs
  4. Network Security: Ensure secure connections to OpenAI API

Advanced Features​

Bulk Operations​

You: Add HTTP monitors for all services in my staging environment:
- api-staging.example.com:8080
- web-staging.example.com:80  
- admin-staging.example.com:3000

Limitations​

  • Requires active internet connection for OpenAI API
  • OpenAI API usage incurs costs based on token consumption
  • Complex multi-step operations may require multiple interactions
  • Some advanced configurations may need manual setup
  • Rate limiting may apply based on OpenAI plan

Support​

For issues with AI Agent Chat:

  1. Check this documentation first
  2. Review application logs for errors
  3. Test OpenAI API connectivity independently
  4. Contact HertzBeat support with specific error messages