A dashboard for monitoring API usage and inference calls.

• User summary view with aggregated usage data
• All requests view with collapsible rows showing inference calls
• Inference calls view with filtering options
• Pagination for handling large datasets
• Comparison between usage data and inference calls for debugging
• Fallback to inference data when usage data is missing

The dashboard is built with a modular architecture:

/usage-dashboard/
  index.ts                 # Main entry point and routing
  auth.ts                  # Authentication logic
  /api/
    index.ts              # API request handler
    requests.ts           # API endpoints for requests data
    inference-calls.ts    # API endpoints for inference calls
    user-summary.ts       # API endpoints for user summary data
  /views/
    layout.ts             # Common layout template
    dashboard.ts          # User summary view
    requests.ts           # Requests view with collapsible rows
    inference-calls.ts    # Inference calls view
  /utils/
    formatters.ts         # Shared formatting utilities
    pagination.ts         # Pagination utilities

Access the dashboard at the root URL. Authentication is required.

1. User Summary - Shows aggregated usage data by user

   • When usage data is missing, it falls back to inference call data
   • Indicated with an "I" badge

2. All Requests - Shows all usage requests with pagination

   • Click on a row to see associated inference calls
   • Compares usage data with inference call totals
   • Highlights discrepancies for debugging

3. Inference Calls - Shows all inference calls with pagination

   • Can be filtered by usage_id
   • Links back to the associated request

• /api/requests - Get paginated requests
• /api/requests?usage_id=123 - Get a specific request
• /api/inference-calls - Get paginated inference calls
• /api/inference-calls?usage_id=123 - Get inference calls for a specific request
• /api/user-summary - Get user summary data

The dashboard includes several features for debugging:

1. Collapsible Rows - Click on a request to see its inference calls
2. Token Comparison - Compare tokens and price between usage and inference data
3. Discrepancy Highlighting - Visual indicators when data doesn't match
4. Inference Data Fallback - Use inference data when usage data is missing