insights/default

insights@1: one view of the running instance. manifest() is the core's kestrel.describe() (modules with version, contracts, config schema and variables, steps with their descriptions, pipelines with step order, triggers) — static, computed once; stats() aggregates the core's run observer in this process: count, failed (every non-ok outcome), errors (5xx), p50/p95 ms per pipeline and per step spec, active runs, uptime, event-triggered runs per event name, and a ring buffer of the last failed runs (recentFailures, newest first: time, run id, pipeline, trigger, status, duration, code, step and the message the caller received — truncated, never a stack). The module hooks in through the core's attach lifecycle hook and detaches on stop(). Numbers are per process and start at zero on every boot. Every config variable in the manifest carries its effective value: value is what the config sets, otherwise the default, null when nothing applies, and status says which of the two it is. The value is a JSON snapshot — functions, class instances and Buffers become type labels, long strings and large arrays/objects are cut with a marker. A secret is never in it: redacted: true with value: null covers everything the schema marks secret and everything whose key name or value looks like a credential, nested keys inside a shown value included ("[redacted]"). A module author marks a secret with .describe("secret") on the zod field (the core's SECRET constant); a config schema written as JSON Schema marks it with writeOnly: true, format: "password" or x-secret: true. Because values are on show, insights.readManifest belongs behind a login and insights.read — as in the example pipelines below. Config: { recentFailures } — how many failed runs to keep (default 50, 0 turns the buffer off). Steps: insights.readManifest (adds generatedAt), insights.readStats; both write result and read nothing — guard them with authn.requireUser and authz.require:insights.read in the pipeline. Not included: routes (the consumer's kestrel.config.ts wires them), persistence or time series (OpenMetrics/Prometheus is the way there), ratelimit buckets (ratelimit stays [] until a ratelimit contract exists), numbers across processes.

Generated from the manifest

@michaelthielemann/kestrel-insights – module insights/default: provides insights@1.

| Config | Type | Required | Default | |---|---|---|---| | recentFailures | integer | no | 50 |

| Step | Summary | Reads | Writes | Input | Output | Errors | |---|---|---|---|---|---|---| | insights.readManifest | The static manifest of this instance: modules, contracts, config schemas, steps, pipelines, triggers | – | result | – | { generatedAt: number, core: object, contracts: string[], modules: object[], steps: object[], pipelines: object[], triggers: object, … } | – | | insights.readStats | Live counters of this process: runs, pipelines, steps, events, and the most recent failed runs | – | result | – | { generatedAt: number, process: object, runs: object, pipelines: object[], steps: object[], events: object[], ratelimit: object[], recentFailures: object[], … } | – |

Pipelines in examples/minimal using these steps:

• insightsManifest (GET /admin/insights/manifest): authn.requireUser → authz.require:insights.read → insights.readManifest
• insightsStats (GET /admin/insights/stats): authn.requireUser → authz.require:insights.read → insights.readStats