What the run measured, or null when it recorded no output.
The corpus request's ID.
The Choice's agent (decision), or the baseline's first row, or null.
The whole discovery's latency (options, search and call), or null when not recorded.
An agent label's agent, or null.
A none label's kind, or null.
For an agent label: whether the labelled agent could be chosen at all.
The decision call's latency, or the baseline's search.
The baseline's best-ranked candidate regardless of the floor, or null.
Whether the discovery finished within production's timeout; null without an answer.
Whether production would suggest the chosen agent, as the driver judged it; null without an answer.
One run of one request in one cell, as the metrics read it.