API reference

Home summary

One request for a workspace's home view - alerts computed from its own records, the last 7 days against the 7 before, recent calls and agents.

The dashboard's Home page reads everything it shows from one endpoint. You can call it too, for example to put the same alerts on your own status board.

Operation	Method and path	Minimum role
Get the home summary	`GET /dashboard/home`	viewer

An API key acts as admin.

Get the home summary

GET /dashboard/home

Query parameter	Meaning
`tz`	The IANA time zone days are counted in, such as `Asia/Kolkata`. Default `UTC`; an unknown zone is a `400`.
`include_simulated`	`true` counts simulated calls in the figures and lists. Default `false`.

bash

curl "https://api.sevennadi.si/dashboard/home?tz=Asia/Kolkata" \
  -H "Authorization: Bearer vk-0123456789abcdef0123456789abcdef"

Response 200 OK

alerts: what needs attention, most severe first. Each has a kind, a severity (high, medium or low), a title, a one-sentence message, and the agent_id and agent_name it is about (or null). Every alert comes from stored data:

`kind`	Raised when
`workflow_tests`	The agent's latest workflow-test run had failed scenarios or failed safety probes (`high`), flaky scenarios (`low`), or stopped with an error. Carries `run_id`.
`simulated_calls`	The agent's latest simulated-call run had calls that failed their checks, or stopped with an error. Carries `run_id`.
`wallet`	The balance is below the low-balance threshold (`medium`) or has run out (`high`).
`safety_steps`	Calls today whose graph path reached a node that one of the agent's safety transitions (`global_edges`) leads to.
`call_errors`	Calls of the agent that ended `failed` or `error` in the last 24 hours (`high` from 5).
`pipeline`	The agent's voice pipeline is not configured, disabled or gone, so its calls fail. Carries `pipeline`.
`engine`	The platform checks its voice engine's health and the check failed. Never raised when the platform has no health check.
`incident`	Only in the platform operator's own workspace: an open incident from the platform's monitoring (a model server or the voice engine unreachable, slow speech-to-text, a high call error rate, a job backlog, low disk space, a stale backup). The title says since when ("Model server unreachable (STT) since 23:45", in `tz`); carries `since` and an `incident` object (`id`, `key`, `kind`, `severity`, `opened_at`). A customer workspace never receives it.
`monitoring`	Only in the operator's workspace: the monitoring itself has stopped reporting (carries `since`).

kpis: current (the last 7 days, today included) and previous (the 7 days before), each with calls, completed, failed, completion_rate (completed out of calls that ended, null when none ended), safety_triggers and spend (call cost in major units of currency).
calls_per_day: the last 7 days, oldest first: date, calls, completed, failed, spend.
recent_calls: the 5 latest calls, with agent_name, status, to_number, conversation_duration, total_cost (minor units) and simulated.
agents: every agent, most recently updated first, with kind (prompt or graph), its pipeline (id, name, status) and calls_7d.
wallet: balance, currency, low_balance_threshold, low.
engine: {"ok": true|false} when the platform checks its voice engine, else null.

The per-call response time is not stored on calls, so it is not part of the figures; simulated-call runs report their own latency.