# Usage Tracking — Real-Time API Consumption Monitoring

Source: https://ofox.io/docs/develop/observability/usage-tracking

The OfoxAI Console provides comprehensive usage statistics and performance monitoring.

## Real-Time Usage Statistics

### Overview Dashboard

The console homepage displays key metrics:

- **Total requests** — Total API calls in the current period
- **Total tokens** — Total prompt + completion tokens consumed
- **Total cost** — Cumulative cost in the current period
- **Success rate** — API request success rate

### Time Trends

View trends across different time dimensions:

| Dimension | Description |
|-----------|-------------|
| Real-time | Last 1 hour, per-minute granularity |
| Daily | Last 24 hours, per-hour granularity |
| Weekly | Last 7 days, per-day granularity |
| Monthly | Last 30 days, per-day granularity |

## Multi-Dimensional Analysis

### By Model

View usage breakdown per model:

- Request count share
- Token consumption share
- Cost share
- Average latency

### By Provider

View performance across provider nodes:

- Success rate comparison
- Latency distribution
- Fallback trigger frequency

### By API Key

View usage per API Key:

- Request volume per key
- Cost per key
- Useful for tracking costs by project/team

## Alerts

### Alert Rules

Configure the following alerts:

| Alert Type | Description |
|-----------|-------------|
| Error rate alert | Error rate exceeds threshold (e.g. 5%) |
| Cost alert | Daily/weekly/monthly cost exceeds budget |
| Latency alert | Average latency exceeds threshold |
| Fallback alert | Abnormal fallback trigger frequency |

### Notification Channels

Alerts support the following notification methods:

- Email notifications
- Webhook (integrates with Slack, Lark, DingTalk, etc.)

Alert configuration is managed in the console under **Settings** → **Alert Rules**.

## Best Practices

1. **Use different API Keys for different projects** — Makes it easier to track costs by project
2. **Set cost alerts** — Avoid unexpected high bills
3. **Regularly review model distribution** — Ensure model selection matches expectations
4. **Monitor error rate trends** — Abnormal increases may indicate configuration issues