"I want to log custom evaluation metrics, and user feedback"

"I want human reviewers to annotate my traces"

"I want to filter and export traces by annotations"

We heard you, and we shipped.

Now, you can...

• Add tags to logged inferences (on the UI, or via the API)
• Create datasets from tagged inferences, and run experiments
• Filter and export logs with specific tags
• Log user_feedback (👍/👎)
• Log custom metrics (for example: dwell_time, click_rate, suggestion_accepted) (docs)

What can you do once you tag your logs?

• Filter and add to a dataset to run experiments or evaluate
• Curate datasets for fine-tuning