verdicts field of write memories). This read answers “what was the verdict on this block, as known at as_of”: one item per requested target_id, with the latest verdict picked here so you do not sort them yourself.
The latest is the row with the greatest (decided_at, captured_at) among rows decided by valid_at (default as_of). When the top two tie on both, the server does not pick: verdict is null and reason is unresolved_conflict, with every row in others. A verdict written after as_of is hidden and counted in masked.verdict; one decided after valid_at is visible but reads not_yet.
Query parameters
Repeat it for several targets, 1 to 1000 per request. More is a
400 (TCKG_TOO_MANY_IDS); none is a 422 (the parameter is required).Knowledge time. Only verdicts known by then. A date is widened to its end of day (UTC). Without it the request is a
400 (TCKG_AS_OF_REQUIRED).Content time. A verdict decided after it reads
not_yet and is not a candidate for verdict. Default as_of.Response
One per distinct
target_id, in id order: target_id, verdict (the latest row, or null), others (every other visible row, newest first, not_yet rows included), reason (null, no_verdict when nothing is visible, or unresolved_conflict).A row of
tckg.read_verdicts: id, target_id, tenant, value and outcome as {status, value} (status is known, not_yet or unknown), method, ruleset, resolver, decided_at, captured_at.Scoped to
space, like every read. masked.verdict counts the verdicts hidden because they were known after as_of.