Home

Home is the landing page of the dashboard. It shows a snapshot of a single agent over a selected time window, with every rate compared against the previous window of the same length. Use it to decide, in one glance, whether anything needs your attention today.

​

Page controls

• Agent selector — scopes the whole page to one agent, matched by the agent name in your telemetry. Conversations without an agent name are grouped under Unlabelled / default.
• Date range — last 24 hours, 7 days, 30 days, or 90 days, or a custom range. Each metric is compared to the window of equal length immediately before it.
• Refresh — reloads the snapshot. The page also refreshes automatically every 5 minutes while the tab is visible.

The selected agent and window are encoded in the URL, so a filtered view can be shared as a link.

​

Agent health

A one-line assessment of the selected agent in the selected window, with a state badge: The Review issues link opens Problems.

​

Agent snapshot

Four tiles summarize the window: Each quality tile links to the matching tab on the Signals page, pre-filtered to the same agent and window. A Health over time chart beside Next actions plots each rate for the current window (solid line) and the previous window (dashed line).

​

Next actions

“What needs review, ranked by impact.” A mixed queue of up to four rows: Each row shows the number of affected conversations and links directly to the problem, signal view, or proposal.

​

Where the agent is breaking

The top failing tools, ranked by affected conversations rather than raw call volume. Each row shows the tool name, affected conversations, failure rate, and failed-call count, and links to the Tool Failures tab filtered to that tool. View tool failures opens the full list.

​

What the agent handles

The top use-case clusters, ranked by linked conversations. A conversation can belong to more than one use case, so rows may overlap. Each row links to the use-case detail; View use cases opens Use Cases. This section is empty until the first use-case taxonomy has been built, which requires at least 50 analyzed conversation segments.

​

Data details

A collapsible footer shows when the snapshot was generated, how fresh each data source is (the timestamp each source has data through), and any sources that were unavailable for this snapshot. Rates keep their source-specific denominators, so the tiles are not directly comparable to each other.

​

Next steps

• Signals — drill into the detections behind each quality rate.
• Problems — work the ranked list of root causes.
• Conversations — inspect individual conversations end to end.
• How Moda analyzes conversations — what Moda computes and where each result appears.