🔸Reviewing Run Analytics

The Run Analytics page is tailored to offer users real-time observability metrics to track their model's performance over time. It features charts that detail a model’s latency, token usage, and the associated costs of running inferences. Users can apply filters for different models and tags to specifically analyze latency and cost related to a particular dataset on a selected model.

1. After assessing the evaluation from the previous step, navigate to the Run Analytics page to see charts on Latency, Cost, Queries, and Tokens for a given time period.

Example screenshot #1 of the Run Analytics page

Example screenshot #2 of the Run Analytics Page

2. To filter on the runs we ran for this particular example, the following filters would need to be applied:

   • Start Date

   • End Date

   • Model: Mistral, OpenGPT

   • Tags: CEOs, Example

The page should look something similar to the following screenshots:

Screenshot #1 of the Run Analytics page, using the example for this walkthrough

Screenshot #2 of the Run Analytics page, using the example for this walkthrough

Hover over the points on the line graph and bar charts to view detailed information! (example seen in Screenshot #2)

3. The data from the runs can be viewed at the bottom of the page. This is the same data that can also be viewed on the Prompt Runs page.