Dashboard

Overview

The Observability dashboard provides real-time insights into your AI applications. Monitor traces, analyze costs, track performance, and debug issues from a single view.

Dashboard Sections

Summary Cards

At the top of the dashboard, key metrics at a glance:

Time Range Selector

Filter data by time period:

• Last hour
• Last 24 hours
• Last 7 days
• Last 30 days
• Custom range

Traces Chart

Visualize trace volume over time:

• Stacked by status (success, error)
• Hover for detailed counts
• Click to drill into specific time periods

Token Usage Chart

Track token consumption:

• Input vs output tokens
• By model
• Trend over time

Cost Chart

Monitor spending:

• Daily/hourly breakdown
• By model comparison
• Cumulative view

Navigating the Dashboard

Filters

Filter data by:

• Model: gpt-4o, claude-3-5-sonnet, etc.
• Status: Success, Error
• User: Specific user IDs
• Tags: Custom tags
• Session: Specific sessions

Search

Full-text search across:

• Trace names
• Input content
• Output content
• Metadata

Quick Actions

• Export: Download data as CSV/JSON
• Share: Generate shareable link
• Alert: Create alert from current view

Dashboard Tabs

Traces Tab

List view of all traces:

Click a trace for detailed view.

Sessions Tab

Group traces by session:

Generations Tab

Filter to LLM calls only:

Models Tab

Aggregate by model:

Trace Detail View

Click any trace to see:

Timeline

Visual representation of the trace:

Trace: chat-completion (2.5s)
├── Span: retrieve-context (500ms)
│   └── Generation: embed-query (200ms)
└── Generation: generate-response (1.8s)

Input Panel

Full input data:

{
  "userMessage": "What is machine learning?",
  "userId": "user-123",
  "sessionId": "session-456"
}

Output Panel

Full output data:

{
  "response": "Machine learning is a subset of AI...",
  "model": "gpt-4o",
  "tokens": 500
}

Metrics

• Start time
• End time
• Duration
• Total tokens
• Total cost
• Status

Metadata

Custom metadata attached to the trace:

{
  "environment": "production",
  "version": "1.2.0",
  "feature": "chat-v2"
}

Customizing the Dashboard

Saved Views

Save custom filter combinations:

1. Apply desired filters
2. Click Save View
3. Name your view
4. Access from Saved Views dropdown

Default Time Range

Set default time range:

1. Go to Settings
2. Under Dashboard Defaults
3. Select preferred time range

Column Visibility

Choose visible columns:

1. Click column selector icon
2. Toggle columns on/off
3. Drag to reorder

Keyboard Shortcuts

Next Steps

• Metrics - Key metrics explained
• Cost Analysis - Detailed cost breakdown
• Performance - Latency analysis