AI Canvas
AI Canvas is an interactive investigation workspace where you ask questions about your system and the assistant lays out its findings on a spatial canvas. Each question starts a new canvas; the assistant's reasoning — incidents queried, monitors checked, time-series data retrieved — appears as cards you can pan and zoom around.
INFO
AI Canvas requires an active license with the AI Canvas feature enabled. If your license does not include AI Canvas, the Canvas entry does not appear in the sidebar. Contact your administrator to upgrade your license.
Getting started
Navigate to Canvas in the sidebar to open the canvas list. Type a question in the prompt and press Ask (or Enter) to start a new canvas. Pulse sends the question to the AI assistant and opens the canvas as the assistant begins working.
You can ask about anything related to your monitored system — for example:
- "Are there any active incidents on the production line?"
- "Show me the signal trend for Sine0 over the last hour."
- "Which monitors are currently unhealthy?"
- "Which of my devices have critical open CVEs?"
The canvas
Once the assistant has returned results, the canvas opens in a split view:
- Left — spatial canvas: The assistant's findings appear as cards arranged in a zoomable, pannable workspace. Each step the assistant took becomes its own card.
- Right — chat panel: The conversation thread shows the question you asked and the assistant's text response.
Cards
Each card on the canvas shows the result of one step the assistant took:
| Card type | What it shows |
|---|---|
| Text | The assistant's narrative explanation or your message |
| PQL chart | A time-series chart from a data query |
| Incident list | Incidents matching the assistant's search |
| Monitor list | Monitors matching the assistant's search |
| CVE finding list | Vulnerability findings matching the assistant's search |
| Markdown | Structured text output from a tool |
Click the </> icon in the top-right corner of any tool card to toggle between the result and the raw tool call (the input the assistant sent to the tool).
Navigating the canvas
| Action | How |
|---|---|
| Pan | Drag empty canvas space |
| Zoom | Ctrl / ⌘ + scroll wheel, or pinch on a trackpad |
| Scroll (pan without zoom) | Scroll wheel (without Ctrl / ⌘) |
| Zoom in / out | + / − buttons in the bottom toolbar |
| Fit all cards to view | ⤢ (fit) button in the bottom toolbar |
The current zoom level is shown as a percentage between the zoom buttons.
Renaming a canvas
Canvas owners can rename a canvas by clicking its title in the header bar. Edit the name and press Enter to save, or Escape to cancel. The assistant may also set an automatic title after completing its first run.
Continuing the investigation
Owners can send follow-up questions using the text field at the bottom of the chat panel. Press Enter to send (or Shift+Enter to insert a new line). The assistant runs again and adds new cards to the canvas.
The text field is hidden for users who are not the canvas owner — non-owners can view the canvas and its findings but cannot extend it.
Managing canvases
The canvas list page shows all canvases you own or can view. Each row shows the canvas name, owner, and when it was last updated.
To delete a canvas, open the row's action menu (⋯) and choose Delete. Confirm the deletion in the dialog. Deletion is permanent — the canvas history and all its findings are removed.
Related documentation
- Query — Build ad-hoc time-series queries on your monitor data directly.
- Monitors — The monitors the assistant can query on your behalf.
- Incidents — The incidents the assistant can search and summarize.
- Vulnerabilities — CVE findings the assistant can query and analyze.