Traces

The Traces page gives you a searchable, filterable list of every trace your application has sent to BrowserStack AI Evals. Click any trace to open a detailed view of the full execution.

Trace List

Trace Detail View

Click any row in the trace list to open the trace detail view.

Observation Tree

The left panel shows the observation tree — the full hierarchy of spans and generations that make up the trace.

Timeline

The top panel shows a toogle timeline to view of all observations plotted against wall-clock time. This lets you see which operations ran sequentially and which ran in parallel.

Hover over any bar in the timeline to see the observation name, start time, and duration.

Detail Panel

When you select a node in the observation tree, the right panel shows:

For traces:

• input — the root input to your application
• output — the final output from your application
• metadata — any custom fields attached by your code
• tags — labels attached at trace creation
• user — the user ID, if set
• session — the session ID, if set
• Custom ID — if a trace was created with a custom ID (see Custom trace IDs), the original string is shown as a Custom ID: <value> badge. Click the badge to copy the value to your clipboard.

For generations (LLM calls):

• input — the full message array sent to the model
• output — the model's response
• model — model name and version
• usage — prompt tokens, completion tokens, total tokens
• cost — estimated cost in USD
• latency — time to first token and total duration
• model_parameters — temperature, max_tokens, and other parameters
• tool_calls — any tool/function calls made during the response
• prompt_name / prompt_version — if a managed prompt was used

For spans (non-LLM steps):

• input — the input to the step
• output — the output from the step
• metadata — custom metadata

Scores Tab

The Scores tab on the trace detail view lists all evaluation scores attached to the trace or its observations.

Each score shows:

• The metric name (e.g., faithfulness, correctness)
• The score value (numeric or categorical)
• The source (evaluator, human, or api)
• An optional comment explaining the score

Distributed Traces

When your application uses distributed tracing, spans from multiple services appear in the same observation tree. The SDK propagates a trace context carrier between services so the dashboard can stitch them together into a single trace view.

In the observation tree, you will see spans from all participating services nested under the same root trace. Each service's LLM calls, spans, and events appear in their execution order, giving you full end-to-end visibility across service boundaries.

Navigating Between Traces

• Use the ↑ K (Navigate Up) and ↓ J (Navigate Down) → buttons at the top of the trace detail view to step through traces in the current filtered list without returning to the trace list.
• Press Escape to close the detail view and return to the list.
• The URL updates as you navigate, so you can bookmark or share a link to any specific trace.