Find and triage work in Flightdeck

DVARA Flightdeck brings operational signals and governance evidence into one role-aware application. Use this workflow to find the affected object, decide whether the problem is control-plane or traffic-facing, and collect evidence without copying secrets.

Enterprise only

Not included in DVARA Open Source.

Find a page or governed object

Press Cmd+K on macOS or Ctrl+K elsewhere, or select the search field in the top bar. Search is case-insensitive and returns only destinations your role can open. After two characters it also searches permitted policies, budget caps, and audit events. Workspace roles see results from their workspace; platform roles see the scope their role allows.

Use the arrow keys to move, Enter to open a result, and Esc to close the palette. If an expected result is absent, check your active workspace and role before treating it as missing data.

Start daily triage from the dashboard

Read the dashboard before opening individual records:

1. Check Gateway connectivity and provider health.
2. Review error rate, p95 latency, and the slowest provider/model aggregates.
3. If priority admission is enabled, check current load and throttling by tier.
4. Review pending MCP and A2A approvals and any cost or usage warning.
5. Open the affected workspace, policy, budget, session, or audit event through global search.

Dashboard cards are signals, not evidence by themselves. Follow a card to its task page and use the object's recorded IDs and timestamps for investigation.

Act on notifications

From 1.10.0 Flightdeck raises a notification for work someone has to act on:

• approvals waiting, and a sensitive action waiting for a second approver,
• a failed audit export,
• budget warnings and hard caps reported by gateways,
• A2A client certificates near expiry,
• a model served with no price, from 1.10.0.

The bell in the top bar appears only while you have an open notification, read or not. A repeat of the same problem groups into the open notification instead of adding another. Platform roles see every workspace's notifications.

The first time a notification is raised, Flightdeck writes one NOTIFICATION_RAISED audit event, which webhooks subscribed to NOTIFICATION receive, and queues one email to each user whose role can act on it. A failed email is retried with backoff. An action-required warning still open after 24 hours becomes critical and is sent again. Turn your own emails off on the Notifications page (/notifications).

From the top bar you can also save a page's filters as a private view and return to it later.

Separate Flightdeck readiness from Gateway health

Open Operations → System Status when live traffic panels are empty or the connection indicator is red. Flightdeck readiness depends on its application state and PostgreSQL, not Gateway reachability. A Gateway outage can therefore leave configuration, stored audit evidence, and other control-plane work available while live provider and traffic data is stale.

Continue with Check System Status and recover from errors for safe error codes, correlation IDs, and deployment recovery links.

Use the in-product help for the current task

Open the user menu and select Help to reach the operator guide at /help. The top-bar Docs link follows the current Flightdeck page and opens the closest public guide. Use Help for the UI task in front of you; use Docs for configuration boundaries, defaults, and end-to-end behavior.

Build a safe evidence bundle

Record the workspace, time window, trace or correlation ID, affected object ID, and the visible status or error code. Export the narrowest relevant session, conversation, usage, or audit window. Never include API keys, provider secrets, OAuth tokens, raw PII, or browser cookies.

For an incident that crosses planes, keep the exports separate and join them by trace, session, conversation, workspace, and timestamp. This preserves each audit chain and makes the evidence easier to review.