Skip to main content
This feature is in Early Access (EA), and might not be visible in your app. To learn more about this program and the features it contains, see the Early access page.
The Assistant in Chronosphere Observability Platform reads your telemetry and configuration to answer questions, and acts on the resource you’re viewing when you ask it to. Use this reference to determine what it can read, the skills it applies to specialized tasks, and the changes it can make.
Generative AI features can produce incorrect results, hallucinate data, and deliver inaccurate analysis. Use generative AI features with care, and independently verify all information produced by generative AI tools before applying it.Certain prompts, data, or other inputs might produce irrelevant content. Don’t rely on generative AI features or responses for any uses that exceed their designed scope.

​
Data the Assistant reads

The Assistant queries your telemetry with the permissions of your account’s role. It can’t read data that your account can’t already access. The following table lists the data the Assistant reads and examples of how you can prompt each source: The Assistant reads these sources in a sandboxed environment that allows it to filter and combine results before responding. When you expand a tool call in the conversation, you see the commands it ran and their output. For more information, see Tool calls.

​
Skills

A skill is a set of instructions the Assistant loads for a specialized task, such as writing a query in a specific language or investigating a signal. The Assistant selects skills on its own based on your prompt. To direct it to a specific skill, type / in the message field and select the skill. Table 2: Assistant skills The following table lists the skills available from the slash command menu:

​
Changes the Assistant can make

The Assistant can change the dashboard or notebook you have open, and can navigate you to another page. It validates each change before applying it and tells you what it changed. For dashboards, it stages the changes so you can review and modify them before saving the updated dashboard.

​
Edit the open dashboard

When you’re viewing a dashboard, the Assistant can read its current state, including unsaved edits, and apply changes such as adding a panel or adjusting a query. The Assistant validates its change against the dashboard schema before applying it. When validation fails, it reports the problem and leaves the dashboard unchanged. After a successful change:
  • The dashboard switches to edit mode.
  • The change is recorded in the dashboard’s edit history as Edited dashboard with assistant, so you can undo it like any manual edit.
  • The change isn’t saved until you review it and click Save.
To start, open a dashboard in edit mode and ask the Assistant what you want to change. For example:

Add a panel to this dashboard showing request rate by status code.

​
Edit the open notebook

When you have a notebook open, the Assistant can read its contents as Markdown and edit its text, visualization panels, and tables. Notebook edits are saved automatically, so review them after the Assistant reports a change. Notebooks retain their version history, and you can restore an edited notebook to a previous state. The Assistant edits a notebook by matching exact text. When the text it’s looking for appears more than once or doesn’t appear at all, rather than editing the wrong section, it rejects the change and reports why.

​
Navigate to a page

The Assistant can open another page in Observability Platform for you, such as a dashboard or an explorer. It navigates only to valid Observability Platform routes and documentation pages, and reports an error rather than following a link elsewhere.

​
Start an investigation

This feature isn’t available to all Chronosphere Observability Platform users and might not be visible in your app. For information about enabling this feature in your environment, contact Chronosphere Support.
The Assistant can start a deep investigation and monitor its progress while you work elsewhere. Investigations are available on tenants that have them enabled. To start an investigation from the Assistant, ask it to investigate an alert, dashboard panel, or symptom. Before starting one, the Assistant confirms with you unless your prompt clearly requested it. After starting, it links to the investigation and offers to monitor it. To start an investigation from a dashboard panel:
  1. Hold the pointer over the panel, and then click the three vertical dots icon.
  2. Click Add to investigation. When an investigation already exists for that panel and time range, click Show investigation instead.
Observability Platform opens the Assistant and starts the investigation. Progress appears as an investigation card in the conversation. To continue a completed investigation, open its report and click Continue investigation, describe the direction you want to pursue in the Look deeper into field, and then click Submit. Observability Platform forks a follow-up investigation and hands it to the Assistant to monitor.

​
Escalate from query generation

Natural language query generation produces a single query in the field you’re editing. When it can’t fulfill a request, its error message includes an Ask Assistant link. Click the link to send the same prompt to the Assistant, which opens in the side panel and retries the request with its full set of capabilities, including metric discovery and the /promql skill.

​
Tenant-specific guidance

Chronosphere can configure instructions that the Assistant applies to every conversation in your tenant, such as your naming conventions, the services your teams own, or the dashboards to prefer for a given question. The Assistant follows these instructions alongside its built-in behavior. These instructions don’t change your account’s permissions or grant access to data your tenant can’t otherwise read. To request them, contact Chronosphere Support.