Skip to content
Repo

Sample

from modal_dojo import Sample

A prompt, response, parsed structure, score, and metadata from one model call.

Attributes

score float

Reward or eval score for this call. Default: 0.0

prompt str

Text sent to the model for this call. Default: ""

response str

Raw model output. Default: ""

parsed_response ParsedResponse | None

Structured parse of response, if available.

metadata dict[str, Any]

Custom metadata attached to this call by a reward or rollout function. Default: {}

trace list[TraceSpan] | None

Execution spans for a sampled subset of rollouts when recording is on. None otherwise.