Wall-clock runtime execution duration in milliseconds.
OptionalendpointPer-endpoint runtime feedback for failed e2e tests.
Populated when failed > 0 and the realize history exposes at least one
operation function the failures could be routed to. Each entry pairs an
endpoint identifier (METHOD:/path) with the LLM-ready digest text that
downstream diagnostics or explicit regeneration tooling can use as
anti-pattern context. The structured failed count is kept alongside the
digest so downstream consumers (archive ledger, verdict baseline, runtime
diagnostics) do not need to re-parse the digest text to learn how many e2e
tests fired against the endpoint.
Absent on green runs and on runs where the bundle builder could not map any
failure back to a realize operation (those failures are still counted in
failed so the run is not silently reported as a partial success).
Number of executed tests whose runtime error field is non-null.
Number of executed tests whose runtime error field is null.
OptionalreasonHarness-level failure classification when runtime setup or execution threw before producing per-operation results.
OptionalresetWhether the runtime reset the database before executing tests.
OptionalsimultaneousConcurrent execution limit used by the runtime.
True when at least one test ran and every operation passed.
Number of generated e2e operation tests executed.
Digest of Test-phase runtime execution for the generated test suite.