Creating Alerts

This guide provides detailed instructions for creating each type of alert, with examples of when each is most useful.

Creating an Alert

Basic Steps

1. Navigate to Alerts page

2. Click Create Alert

3. Fill in the alert details:

   • Name - A descriptive name (required, min 3 characters)
   • Description - Optional explanation of what this alert monitors
   • Alert Type - Select from the six available types
   • Threshold - The value that triggers the alert
   • Period - Time window for the alert (hour, day, week, month)
   • Notification Channels - Dashboard and/or email

4. Click Create Alert

Cost Threshold Alerts

Monitor spending and stay within budget.

Configuration

When Triggers

Alert fires when your total spending in the period exceeds the threshold.

Example Scenarios

Daily Budget Control

Name: Daily Cost Alert Threshold: $25 Period: Day Use case: Prevent runaway costs from bugs or unexpected usage

Weekly Budget Monitoring

Name: Weekly Spend Warning Threshold: $100 Period: Week Use case: Track weekly spending against your budget

Monthly Cost Cap

Name: Monthly Budget Alert Threshold: $500 Period: Month Use case: Ensure you don't exceed monthly allocation

Recommended Settings

Error Rate Alerts

Catch API issues before they impact users.

Configuration

When Triggers

Alert fires when error rate exceeds threshold percentage within the period, only if minimum request count is met.

Example Scenarios

API Health Monitoring

Name: High Error Rate Alert Threshold: 5% Period: Hour Min Requests: 10 Use case: Detect when API reliability drops

Production Quality Gate

Name: Critical Error Alert Threshold: 10% Period: Hour Min Requests: 5 Use case: Alert on severe issues affecting many requests

Low-Volume Monitoring

Name: Any Errors Alert Threshold: 1% Period: Day Min Requests: 100 Use case: Catch issues in lower-volume applications

Understanding Error Rates

• Errors include: timeouts, rate limits, server errors, invalid responses
• Normal rate: 0-2% depending on use case
• Warning level: 3-5%
• Critical level: >5%

Usage Spike Alerts

Detect unusual increases in request volume.

Configuration

When Triggers

Alert fires when current period requests exceed baseline by the threshold percentage.

Example Scenarios

Runaway Script Detection

Name: Traffic Spike Alert Threshold: 200% Baseline: Day Comparison: Hour Use case: Detect scripts or bots making excessive requests

Viral Traffic Alert

Name: Unexpected Volume Threshold: 500% Baseline: Week Comparison: Day Use case: Catch unexpected traffic surges

Gradual Growth Tracking

Name: Usage Growth Alert Threshold: 150% Baseline: Month Comparison: Week Use case: Track organic growth in usage

Threshold Guidelines

Quota Limit Alerts

Get warned before running out of credits.

Configuration

When Triggers

Alert fires when the specified percentage of your quota is consumed.

Example Scenarios

Credit Balance Warning

Name: Low Credits Alert Threshold: 80% Quota Type: Credits Use case: Time to purchase more credits

Critical Balance Alert

Name: Credits Nearly Depleted Threshold: 95% Quota Type: Credits Use case: Urgent - service may be interrupted

Comfortable Buffer

Name: Credits Below 50% Threshold: 50% Quota Type: Credits Use case: Early warning for budget planning

Recommended Thresholds

Model Performance Alerts

Monitor response times and API performance.

Configuration

When Triggers

Alert fires when average latency in the period exceeds threshold.

Example Scenarios

General Performance

Name: Slow Response Alert Metric: Latency Threshold: 10000ms (10 seconds) Period: Hour Use case: Catch general slowdowns

Real-Time Applications

Name: Latency Critical Metric: Latency Threshold: 5000ms (5 seconds) Period: Hour Use case: For applications needing fast responses

Batch Processing

Name: Batch Job Slowdown Metric: Latency Threshold: 30000ms (30 seconds) Period: Day Use case: For background processing that can tolerate delays

Latency Expectations by Model Type

Unusual Activity Alerts

Detect anomalies and potential security issues.

Configuration

This alert type uses automatic pattern detection. No specific threshold is required.

When Triggers

Alert fires when the system detects:

• Unusual request patterns
• Requests from new sources
• Abnormal time-of-day activity
• Suspicious content patterns

Example Scenarios

Security Monitoring

Name: Suspicious Activity Use case: Catch potential credential misuse

Anomaly Detection

Name: Unusual Patterns Use case: Identify bot traffic or attacks

What Gets Flagged

• Requests at unusual times
• New IP addresses or regions
• Rapid request pattern changes
• Unusual model access patterns

Notification Channel Selection

Dashboard Only

Best for:

• Non-critical alerts
• Frequent notifications
• When email would be too noisy

Email Only

Best for:

• Critical alerts you can't miss
• After-hours monitoring
• Important threshold breaches

Both Dashboard and Email

Best for:

• Critical cost and quota alerts
• Production error monitoring
• Anything requiring immediate action

Tips for Effective Alerts

Start Conservative

Begin with higher thresholds and lower them as you understand your usage patterns.

Name Clearly

Use descriptive names that tell you what the alert means at a glance:

• Good: "Daily Spend Over $50"
• Bad: "Cost Alert"

Document Purpose

Use the description field to explain why this alert exists and what action to take.

Review and Adjust

After a week of operation, review your alerts:

• Are you getting too many notifications?
• Did you miss any important events?
• Should thresholds be adjusted?

Related Topics

• Managing Alerts - History and acknowledgment
• Alerts Overview - Introduction to alerting
• Usage Analytics - Understand your usage patterns