Dashboard

View: Dashboard The dashboard showing hero cards, timeframe selector, usage tables, and recent events.

The Dashboard is the first entry of the Overview cluster (/dashboard) and the screen admins, tenant admins, and KI-Managers land on after signing in. It provides a live operations view of gateway activity: request volume, cost, guardrail hits, recent traffic, and recent blocked events.

💡 Note: A brand-new workspace with no traffic yet shows a "No data yet" empty state instead of empty charts, until the first requests are recorded.

Timeframe selector

The top of the hero section contains a row of timeframe tab buttons. Selecting a timeframe updates the values of the hero cards and the sparklines for that window.

Option	Window
Today	From your local calendar midnight to now
Yesterday	The previous calendar day
Last 7 days	Rolling 7-day window
Last hour	Rolling 60-minute window
Last minute	Rolling 60-second window

Hero cards

Three cards summarise the selected timeframe at a glance. Each card includes a sparkline chart of the underlying timeseries.

Card	Primary metric	Sub-metric
Requests	Total request count	Cache hit rate (%)
Cost	Cumulative cost	Cost saved via cached responses
Guardrail Hits	Total guardrail activations	Breakdown: blocked · scrubbed · flagged

💡 Note: The dashboard shows costs in your Display currency preference — euros or US dollars — set under Settings › Preferences. Amounts are metered in US dollars; euro figures are an approximate conversion at the daily reference rate. Cost analytics and the Live monitor have their own Currency selector in the page header.

Input Token Caching Efficiency

This card is shown below the main hero cards only when the selected timeframe has both cache-read and cache-write activity (a prompt-caching provider such as Anthropic was used). A period with only non-cached input — for example local myra-fleet models, which do not cache — hides the card rather than showing a 0 % hit rate.

Section	Description
Hit rate bar	Percentage of input tokens served from cache, colour-coded: green (≥ 60 %), amber (≥ 30 %), red (< 30 %)
Served from cache (reads)	Token count, percentage of total input tokens, and cost at the 0.1× cache-read rate
Processed fresh (cache writes)	Token count, percentage of total input tokens, and cost at the 1.25× cache-write rate
New tokens (not cached)	Token count, percentage of total input tokens, and cost at the standard 1× input rate
Cost summary	Total cached input cost · total uncached input cost · compaction savings (if any)

Usage by tenant

The Usage by Tenant table shows per-tenant activity for the selected timeframe. It is visible only when at least one request exists in the period, and is hidden from users without a management or admin role (the member, viewer, ki_manager, and finance-only or unroled users).

Column	Description
Tenant	Tenant slug
Requests	Total requests
Input Tokens	Total input tokens consumed
Output Tokens	Total output tokens consumed
Cost	Total cost

Top models

The Top Models table lists up to five models ordered by request count for the selected timeframe.

Column	Description
Provider	Provider slug
Model	Model name
Requests	Total requests to this model
Cost	Total cost
Avg Latency	Average end-to-end latency in milliseconds

Recent requests

The Recent Requests table shows the last 10 requests across all gateways.

Column	Description
Time	Timestamp
Tenant	Tenant slug
Gateway	Gateway slug
Provider	Provider that handled the request
Model	Model name
Status	HTTP status code badge
Tokens	Input and output token counts (`in+out`)
Cost	Estimated cost
Latency	End-to-end latency in milliseconds
Flags	`cached`, `blocked`, or `rate limited` badges where applicable

Recent guardrail events

The Recent Guardrail Events table shows the most recent blocked requests. It is visible only when blocked events exist, and is hidden from users without a management or admin role (the member, viewer, ki_manager, and finance-only or unroled users).

Column	Description
Time	Timestamp
Tenant	Tenant slug
Gateway	Gateway slug
Outcome	The guardrail outcome
Detector	Detector that produced the outcome
Reason	Block reason string
Latency	End-to-end request latency in milliseconds (not the guardrail-evaluation time; the Dashboard renders the request latency in this column)

💡 Note: The dashboard does not auto-refresh; reload the page to pull the latest data. Use the Export CSV or Export XLSX button at the top right to download the current view as a .csv file or an .xlsx workbook.

Member reactivation

Platform admin only. When you hold the platform admin role, the dashboard shows a member-reactivation roll-up card below the hero cards. The card reports the reactivation e-mails sent in the last 7 and 30 days, how many members reactivated, the opt-outs, and the ten most-mailed workspaces. The same card appears for a single tenant at the top of Cost analytics, where it is visible to admin and tenant_admin. See Reactivation e-mails for the cadence and the per-tenant controls.