Rovo Specialist

The GrafanaSight Rovo Specialist helps teams ask Confluence-native questions about Grafana context. It answers from GrafanaSight’s cached dashboard, alert, annotation, and sync data, and it should mention freshness so responders know whether the answer is current enough for their decision.

Open The Specialist

Open Rovo from a Confluence page in a configured space, then use the agent switcher to select GrafanaSight Specialist. Do not judge GrafanaSight action behavior from the generic Rovo agent.

Verified Prompts

These prompts were tested against the Supabase Production Health Runbook page in the configured SparrowOps Marketing space.

• “Which dashboards in this space mention Supabase?”
• “Summarize active alerts by severity.”
• “Refresh the GrafanaSight cache for this space.”
• “When was GrafanaSight last synced?”
• “Draft an incident summary for a dashboard named Definitely Not A Real GrafanaSight Dashboard using cached data only.”

The agent also ships these conversation starters:

• Are there any firing alerts in prod?
• Show me dashboards tagged payments.
• What dashboards do we have for checkout?
• How fresh is the GrafanaSight cache?

Conversation starters use general examples. Adapt them to the dashboards, tags, services, and folders in your own Grafana cache.

How The Specialist Uses Data

The Specialist is cache-first. It does not need to make a live Grafana request for every question. This keeps responses fast and keeps answers aligned with the same data that Confluence pages and macros use.

It can use:

• Cached dashboard catalog data.
• Cached dashboard detail and panel metadata.
• Cached alert summaries.
• Cached annotations.
• Sync status and freshness.

Prompt Results

The freshness prompt returned the cache health, last successful sync time, next scheduled sync time, synchronized dashboard count, synchronized alert count, and Grafana instance URL.

The alert prompt grouped cached active alerts by severity and included the cache timestamp used for the answer.

The dashboard prompt found the cached Supabase dashboard, folder, tag, panel count, and related alert context.

The negative groundedness prompt did not invent the requested dashboard. It said the dashboard was not found in the cached catalog, then separated that from general cached system-health context.

The refresh prompt opened the Refresh GrafanaSight cache tool confirmation with target all. After confirmation, Forge logs recorded rovo.refresh_data.finished with ok: true and syncResult: success for the same space.

Available Rovo Actions

Action	What it does	Inputs
Query cached dashboards	Searches GrafanaSight’s cached dashboard catalog and dashboard detail metadata without a live Grafana API call.	Search, topic, folder, tag, dashboard UID, limit.
Query cached alerts	Reads cached active alert summaries.	Alert state, severity, label query, search text, limit.
Refresh GrafanaSight cache	Triggers a cache refresh from live Grafana APIs.	Refresh target: catalog, alerts, or all.
Generate incident summary draft	Drafts an incident summary from cached dashboard metadata, active alerts, and annotation snapshots.	Dashboard UID, search text, service, label query, limit.

The action names above are the names exposed by the app manifest. The Specialist should still answer conversationally, but these actions explain what data it can actually use.

Freshness Matters

Use the configuration page or ask the Specialist about sync status before relying on answers during an incident. If the cache is stale, refresh the cache and ask again.

What It Does Not Replace

The Specialist is not a replacement for Grafana during live incident command. Use Grafana directly when you need raw metric exploration, query editing, dashboard changes, or datasource administration. Use GrafanaSight when you want Confluence pages and runbooks to carry the right observability context.