Get a v2 conversation with recent turns
Authorizations
Path Parameters
Response
Successful Response
queued, ringing, active, paused, waiting_for_human, transferred, awaiting_approval, pending_dispatch, closed, expired, error, superseded low, normal, high, urgent live, test, simulation pending, synced, skipped, suppressed Rollup of eval runs that targeted a session.
succeeded = runs where status='completed' AND every score has passed=true.
failed = total - succeeded - pending (failed-status runs + completed-but-not-all-passed).
pending = runs still queued or running.
The critical_* counterparts restrict the same aggregation to runs whose
binding was marked is_critical=True (snapshotted onto the run). They drive
the session headline pill — non-critical evals still run and display but
don't move the verdict. eval_status collapses the critical counts to a
single headline label (none when no criticals are configured).
Resolved, immutable snapshot persisted on a test conversation.