The cell's configuration.
Optionalbaseline?: "semantic-search" | nullRuns this baseline instead of the decision: no decision call is made.
Optionalfailover?: boolean | nullWhether the runner may fail over. Defaults to false when the cell pins a model or vendor.
OptionalmodelId?: string | nullPins the MJ: AI Models row that must answer.
The oracles to run.
OptionalpromptName?: string | nullThe decision prompt. Defaults to Default Decision, the prompt production's discovery asks.
Optionalseed?: number | nullRequested sampling seed. Recorded, not applied: see the driver.
Optionaltemperature?: number | nullRequested sampling temperature. Recorded, not applied: see the driver.
OptionalvendorId?: string | nullPins the vendor.
The request as production reads it (DiscoveryDecisionState).
The Choice's options (BuildDecisionDiscoveryOptionSet).
The decision prompt.
The run's context, for its user.
Bounds the model call.
The runner's parameters for one discovery call: the prompt, the context user, the request as the state, production's questions over the options, a time bound, and the cell's pinning as
override. The prompt run is not attributed to the conversation manager (production setsagentId, which only labels the row), so eval runs stay out of its analytics.