Tracing
Overview of the tracing UI in BrowserStack AI Evals — view, filter, and analyze LLM and Tool Call traces from the dashboard.
Tracing
The Tracing section of the BrowserStack AI Evals dashboard gives you visibility into every LLM call your application makes. Traces are captured by the SDK and appear in the dashboard within seconds.
What You Can Do
View traces Browse the full list of traces captured from your application. Each row shows the trace name, timestamp, model, token usage, latency, and estimated cost at a glance.
Inspect trace details Open any trace to see the complete execution timeline: the observation tree (spans and generations), full input/output messages, token counts, scores, and any metadata your code attached.
Browse sessions Group related traces into sessions to track multi-turn conversations. The session view shows the full sequence of traces within a conversation along with session-level metrics.
Filter and search Narrow down traces by name, user ID, session ID, tags, metadata fields, environment, or date range. Save filters as presets to quickly return to common views.
Distributed tracing When your AI pipeline spans multiple services, the SDK links traces across service boundaries so they appear as a single trace in the dashboard. Spans from all services are merged into one observation tree, giving you end-to-end visibility. See Distributed Tracing & Context Propagation for setup instructions.
Pages in This Section
Traces
Browse the trace list and inspect individual trace details including the observation tree, input/output, and scores.
Sessions
View multi-turn conversations grouped as sessions with per-session metrics.
Filtering
Filter and search traces by name, user, tags, metadata, environment, and date range.
SDK Setup
To send traces to the dashboard, instrument your application with the SDK. See the SDK Tracing section for setup instructions.