observability

List accumulating failure categories for a metric

Changed on

Return the persistent failure categories for a project (optionally one metric), each with a call count and example call ids. Filter by date_from / date_to (ISO datetimes for an exact window, or bare dates for whole-day bounds) to count only assignments in that range, and by agent_ids (comma-separated agent ids) or agent_version_ids (comma-separated agent version ids) to count only calls handled by those agents / agent versions; results are sorted by count descending. Each category also carries agents: the agents that handled its calls, busiest first, each with a call count, its own example call ids, and whether Improve Agent can edit it. And agent_versions: the agent versions those calls ran on, NEWEST VERSION FIRST, each with the version's name, when it went live, a call count, and its own example call ids — read it to tell whether a version introduced or fixed the failure mode.

get/observability/v1/metric-failure-mode-insights/categories/

Request

  • Base URL
    https://api.cekura.ai
  • URL
    https://api.cekura.ai/observability/v1/metric-failure-mode-insights/categories/
  • Auth
    • API key in header X-CEKURA-API-KEY
    • or HTTP bearer
    • or HTTP bearer
    • or HTTP bearer

Query parameters

agent_idsstring

Comma-separated agent ids; only calls handled by these agents are counted.

agent_version_idsstring

Comma-separated agent version ids (the history_ids reported in agent_versions); only calls that ran on those versions are counted.

projectinteger

Tenant scope. Supply one of project when authenticating with a user session or OAuth bearer token; omit when using an API key already scoped to the tenant.

Response

idinteger
projectinteger
metricinteger nullable
predefined_metricinteger nullable
metric_namestring
is_triageboolean

Derived live from the metric's current Insights opt-in + failure criteria (not persisted) — the FE renders a metric's insight card only when this is true.

window_startstring date-time
window_endstring date-time
num_calls_analyzedinteger
total_failures_in_windowinteger
status'pending' | 'running' | 'succeeded' | 'failed' | 'skipped_no_failures'
  • pending - Pending
  • running - Running
  • succeeded - Succeeded
  • failed - Failed
  • skipped_no_failures - Skipped No Failures
headlinestring
overall_status_emojistring
failure_modesunknown
error_messagestring
duration_secondsnumber double nullable
created_atstring date-time
updated_atstring date-time

Changes