Sign in
← AI PulseSep 16, 2026

Deep · news · Multi-source brief

Vercel AI SDK Adds TypeSafe Provider for Experimental Evaluation

The Vercel AI SDK has introduced a new TypeSafe provider for experimental_evaluate, enabling native support for Choice, Score, and Boolean questions within a single request.

By Illumora Editorial

Source · Sep 16, 2026, 6:26 PM · On Illumora · Sep 16, 2026, 6:39 PM

Media from the primary source — shown here so you can stay on Illumora.

Synthesized from multiple allowlisted primaries on the same event. Lanes →

Brief drafted by Illumora’s editorial model from the linked primary source. Ops desk reviews flagged pieces. How we write →

Read the source →Vercel AI SDK Changelog — Release @ai-sdk/harness-pi@1.0.115 · vercel/ai
Save

Vercel AI SDK changelogs indicate the release of @ai-sdk/typesafe-ai@3.0.0, which includes a new TypeSafe provider for experimental_evaluate. This update supports native Choice, Score, and Boolean questions in a single request, along with structured JSON rubrics, probability distributions, confidence metadata, and workflow serialization.

Key Points

  • The TypeSafe provider for experimental_evaluate is a major change in @ai-sdk/typesafe-ai@3.0.0.
  • It supports native Choice, Score, and Boolean question types within a single evaluation request.
  • The provider includes structured JSON rubrics for evaluation.
  • It also supports probability distributions and confidence metadata.
  • Workflow serialization is a feature of the new TypeSafe provider.
  • The ai@7.0.103 package received a patch to add mid-conversation tool discovery and updates for code-mode.
  • Another patch for ai@7.0.103 deprecates rawInput in output-error UI message parts.

Context

According to the Vercel AI SDK changelogs, the TypeSafe provider is designed to enhance the experimental_evaluate functionality by offering more structured and comprehensive evaluation capabilities. This includes handling various question formats and providing detailed metadata for evaluation results. The updates also address underlying infrastructure concerns, such as protecting Node.js downloads with DNS validation and connection pinning when frameworks wrap global fetch, as noted in the changelogs.

Why It Matters

This update provides builders with more robust tools for evaluating AI model outputs, particularly for structured assessments. The ability to define evaluation criteria using structured JSON rubrics and handle different question types natively can streamline the development and testing of AI applications, potentially leading to more reliable and consistent model performance assessments.

What To Do

  • Note the availability of the TypeSafe provider for experimental_evaluate in @ai-sdk/typesafe-ai@3.0.0`.
  • Review the documentation for experimental_evaluate to understand how to implement native Choice, Score, and Boolean questions.
  • Explore the use of structured JSON rubrics for defining evaluation criteria in your projects.
  • Consider how probability distributions and confidence metadata can inform your model evaluation processes.

Keep Exploring

/techniques/output-schema