Opening the Analytics tab
- Navigate to the expert detail page at
/experts/<id>(or open the expert from the marketplace and click its name). - Select the Analytics tab in the tab strip near the top of the page.
Run list
The run list displays one row per invocation. Each row includes:| Column | Description |
|---|---|
| Request ID | Unique identifier for the invocation (UUID assigned at invoke time). |
| Status | Current or final status of the run, shown as a color-coded badge. See Status badges for all values. |
| Started at | Timestamp when the invocation began. |
| Duration | Wall-clock time from start to completion or failure. |
| View | Action button that opens the run detail drawer. Only shown for invocations that have a recorded request ID; older runs without one cannot be opened from this list. |
Status badges
Each run is labeled with a status badge indicating its outcome. The possible values are:| Status | Meaning |
|---|---|
queued | The invocation has been accepted and is waiting to be picked up. |
running | The expert is actively processing the request. |
succeeded | The invocation completed successfully and produced output. |
failed | The invocation ended with an error. Open the run detail drawer for the error message. |
cancelled | The invocation was cancelled before completion (e.g. via the cancel endpoint or a timeout). |
A run that times out at the platform level is recorded as
failed with an explanatory error message in the run detail drawer.Run detail drawer
Click View on any row in the run list to open the Run detail drawer on the right side of the page. The drawer displays full metadata for that invocation:- Request ID and status badge
- Started-at timestamp and duration
- Input message sent to the expert
- Output produced by the expert (if the run succeeded)
- Error message (if the run failed)
- Thread token (if the run was part of a multi-turn conversation)
If the trace contains more child runs than can be loaded into the timeline, the drawer shows the message: “Trace has more child runs than we could load — timeline shows the first 300 only.” Only the first 300 child runs are rendered; deeper subtrees are truncated.
Next steps
Talking to Your Expert
Send messages and manage conversations with an expert.
API Reference — Invoke an Expert
Invoke experts programmatically and retrieve run status via the REST API.