DecisionRecord
What one tag() call asked and what came back, as plain JSON. Save it,
add expected, and score the tagger with tagger.evaluate().
Fields
- formatrequired
1 The record's format. Records in other formats are skipped.
- idoptional
string The request's
id.- createdAtrequired
string ISO 8601.
- imagerequired
{ sha256: string; mimeType: ImageMimeType; bytes: number } Identifies the image without storing it.
- contextrequired
string As sent, including the default when none was given.
- questionsrequired
readonly Question[] As sent, including default wording and any none-of-these option.
- answersrequired
DecisionResponse["answers"] Raw probabilities by question ID, then option key.
- modeloptional
string The backend's model name.
- modelVersionoptional
string The exact model that answered, when the backend reports it.
- expectedoptional
Readonly<Record<string, ExpectedValue>> The right answers by category, added by you.
Used by
image-tagger is a Ben Hodgson projectGitHub