AI agent main dashboard
The AI agent main dashboard shows what your AI agents are doing right now on a single screen. From top to bottom it is arranged as total counts for the range → flows currently running → recently finished transactions, so you can gauge the scale first and then drill down to individual calls.
Go to Dashboard > AI agent main dashboard in the left sidebar.
This dashboard is read-only. You cannot add widgets or change their positions. To compose the widget layout yourself, use the Overview dashboard. You can look up each widget placed here in the Widget reference.
Screen layout
| Area | Widget | Description |
|---|---|---|
| ① | Option bar | Sets the query time and the application filter. |
| ② | Three summary counts | Shows the total trajectory, LLM API call, and tool call counts for the query range. |
| ③ | Active transaction speed · Hitmap | Shows the speed of transactions in progress and the response time distribution of completed transactions. |
| ④ | LLM active status · Active trajectories | Shows the LLM calls and trajectories currently running. |
| ⑤ | Real-time prompts | Shows the 50 most recently finished LLM transactions in a table, together with their prompts. |
Option bar
Set the query time and the application filter in the option bar at the top.
- The query range is the most recent 10 minutes based on the selected time.
- In real-time mode the range moves forward every 5 seconds and the widgets refresh with it. If you specify a past time, the 10-minute range at that point is fixed.
- You can narrow the targets with the application filter. After changing the filter, click the Apply button for it to take effect.
Summary counts
This area shows the total counts for the query range as one line of large numbers. Clicking a number opens a modal with a trend chart that splits the same range into 5-second intervals.
| Widget | Value shown | Secondary value |
|---|---|---|
| Trajectory count | Number of trajectories newly started during the range (based on mtid) | Number of trajectories currently running |
| LLM API call count | Number of LLM API calls that occurred during the range | Average number of calls per trajectory |
| Tool call count | Number of tool calls (tool use) that occurred during the range | Ratio (%) against all LLM API calls |
The large number is the count started within the range, while the secondary value of the trajectory count is a snapshot of the current moment. The reference points differ, so do not interpret the two values by adding or subtracting them.
The unit for counting trajectories is not a transaction but one flow grouped by the same mtid. Even if several transactions are grouped together or the flow passes through several processes, it is counted as one. The value is the sum of the start counts the SDK sends every 5 seconds over the query range, so trajectories that start and end within 5 seconds are not missed either.
For a detailed description of each widget, see Widget reference — Summary counts.
Flows currently running
- Active transaction speed — Shows requests per second (RPS), responses per second (TPS), and the number of transactions in progress in real time.
- Hitmap — Plots the response time of completed transactions as points on the time axis (X) and the elapsed time axis (Y). The higher a point, the slower the transaction.
- LLM active status — Shows the LLM calls currently being processed, separated by type.
- Active trajectories — Shows the number of trajectories in progress and their distribution by grade. Clicking a grade opens the list of those trajectories. For details, see Widget reference — Active trajectories.
Real-time prompts
This table occupies a whole row at the bottom of the screen. It shows the 50 LLM transactions with the most recent end times, together with their prompts, and refreshes every 5 seconds.
- The Prompt column shows the input of the first step and the output of the last step of that transaction in one cell, along with the evaluation result if there is one.
- Clicking a prompt opens the LLM trace details of that transaction in a new window.
- The only sort criterion is the end time. Transactions that ended with an error also appear, and in that case the response, tokens, and cost may be empty.
For details, see Widget reference — Real-time prompts.