The Dashboard is the Web UI home screen and the instance’s quick “cockpit.”
It shows how busy and healthy your AI employees are right now and links to the
details. The screen loads itself and refreshes every 10 seconds.
What it is for
Open the Dashboard first after signing in: it shows whether conversations are
flowing, whether latency has grown, what share of turns succeed, and whether the
host (gateway, database, channels) is alive. It is the entry point for diagnosis —
every card links to a deeper section.
What it shows
Key metrics (4 tiles): Sessions (number of conversations), Turns
over 24h, p95 latency (95th percentile response time) and success rate
(succeeded turns over total). Missing values render an honest “—”, never a
fabricated number.
Resource tiles — cross-domain counts (agents, plugins, conversations,
schedules, etc.). The same numbers drive the sidebar count badges.
Recent sessions — a card of recent conversations with an “All →” button to
the Sessions screen.
Host health — gateway and database liveness plus channel status; the button
opens the “Status” screen.
How to use it
Watch success rate and p95: a dip is the first signal of a model,
channel or tool problem.
Jump from a resource tile to its section if a count looks unexpected (for
example, zero active agents).
The host-health card is the fast way to tell that “everything is broken” because
of a downed sidecar or channel, not a specific agent.
Configuration
The Dashboard has no settings of its own — it aggregates analytics, conversations,
overview counts and onboarding status. Explore metric windows on the “Analytics”
screen, and load thresholds via Budgets & limits.
The **Dashboard** is the Web UI home screen and the instance's quick "cockpit."
It shows how busy and healthy your AI employees are right now and links to the
details. The screen loads itself and **refreshes every 10 seconds**.
## What it is for
Open the Dashboard first after signing in: it shows whether conversations are
flowing, whether latency has grown, what share of turns succeed, and whether the
host (gateway, database, channels) is alive. It is the entry point for diagnosis —
every card links to a deeper section.
## What it shows
- **Key metrics** (4 tiles): **Sessions** (number of conversations), **Turns**
over 24h, **p95 latency** (95th percentile response time) and **success rate**
(succeeded turns over total). Missing values render an honest "—", never a
fabricated number.
- **Resource tiles** — cross-domain counts (agents, plugins, conversations,
schedules, etc.). The same numbers drive the sidebar count badges.
- **Recent sessions** — a card of recent conversations with an "All →" button to
the [Sessions](/en/v1.0/webui/sessions) screen.
- **Host health** — gateway and database liveness plus channel status; the button
opens the "Status" screen.
## How to use it
- Watch **success rate** and **p95**: a dip is the first signal of a model,
channel or tool problem.
- Jump from a resource tile to its section if a count looks unexpected (for
example, zero active agents).
- The host-health card is the fast way to tell that "everything is broken" because
of a downed sidecar or channel, not a specific agent.
## Configuration
The Dashboard has no settings of its own — it aggregates analytics, conversations,
overview counts and onboarding status. Explore metric windows on the "Analytics"
screen, and load thresholds via [Budgets & limits](/en/v1.0/webui/budgets).
## Next
- [Sessions](/en/v1.0/webui/sessions) — read and run conversations.
- [Budgets & limits](/en/v1.0/webui/budgets) — watch token and money spend.
- [Observability](/en/v1.0/operations/observability) — metrics and logs in depth.
