@ai-sdk/openai@4.0.69
·
33 commits
to main
since this release
Patch Changes
- d4d96bf: Add experimental Choice, Score, and Boolean evaluations through
openai.evaluationModel()and a shared structured language-model evaluation adapter in@ai-sdk/provider-utils/experimental-evaluation. Preserve exact labels and metadata, validate score bounds, and return prompted Boolean P(true) estimates validated to be in [0, 1]. Boolean estimates are not guaranteed to be calibrated; application code chooses thresholds. - Updated dependencies [215b25e]
- Updated dependencies [d4d96bf]
- Updated dependencies [a7dd893]
- Updated dependencies [3456e2c]
- Updated dependencies [c4e76de]
- @ai-sdk/provider-utils@5.0.43
- @ai-sdk/provider@4.0.17