Unify Logo Footer.svg
AI Solutions
Logo
Solution Trace

Solution Trace

Logo

2 mins READ

The Trace tab is the detailed record of what the Solution agent did in a chat — each step it took, how long it took, and what it cost. It's mainly useful for troubleshooting.

Overview

Trace becomes available once a chat has started. It shows the traces for the chat you have open — the agent's planning hand-offs, plan writes, object and workflow changes and other tool calls. For recently built solutions it also includes the app's build steps.

The header shows the number of traces and totals for the session: latency, input tokens, output tokens and cost.

Note: For older solutions, the totals may cover only the conversation and not the app build.

Reading a Trace

  • The list on the left shows the session's traces. Select one to see its timeline on the right.

  • Select a span in the timeline to see its details — "Select a span to see its inputs and outputs" — under Attributes, Events, Input and Output. Switch between Default and JSON views.

  • Export downloads the session's traces.

  • Refresh re-reads the open trace.

Timing

Traces are recorded in the background, so a step that just finished can take a few minutes to appear. If you see "No traces yet — spans take a few minutes to appear after a run", wait a moment and click Refresh. The trace list and session totals refresh periodically while the session is open. The timeline for an opened trace does not refresh automatically; click Refresh to reload it.

Traces and Chats

Traces belong to a chat. Switching to another chat shows that chat's traces. To see where the agent's simpler step summary appears, see Chatting with the Solution Agent.

FAQs

Why is the Trace tab empty right after the Solution agent finishes?

Traces are recorded in the background and take a few minutes to appear. Wait a moment, then click Refresh.

Can I see how much a solution's chat cost to run?

Yes. The Trace tab's header shows the session's total cost, along with latency and input and output tokens.

What does a trace record?

A trace is made up of spans — model calls, tool calls, retrieval steps and other operations the agent performed. Each span shows its timing, inputs, outputs, tokens and cost.