New monitor page for easy analytics

We're launching a new monitor page that provides comprehensive performance and usage patterns across both logs and experiments. This tool is useful for:

• Debugging performance issues by correlating different metrics
• Optimizing costs by identifying expensive patterns
• Tracking the impact of prompt changes on model performance
• Monitoring production stability through response times

The monitor page aggregates the key metrics that matter for LLM applications:

• Latency
• Token counts
• Time to first token
• Cost
• Request volume
• Model performance scores

What makes this particularly powerful is the ability to analyze these metrics across both your logs and experiments, giving you a complete picture of your application's behavior in both production and testing environments.

Flexible data analysis

Understanding that LLM applications often have complex metadata structures, we've built the monitor page with flexibility in mind. You can:

• Group your data by any metadata field, including custom fields you've defined

• Apply filters to individual charts to focus on specific segments of your data

• Analyze data across different timeframes to understand trends and patterns

• Drill down into specific datapoints to view the corresponding traces

Getting started

The monitor page is now available in all projects. To begin using it, just navigate to Monitor in the top navigation of any project. If you have requests for other metrics you'd like to see, let us know!