Sign in
← AI PulseSep 16, 2026

Wire · news · Multi-source brief

Vercel AI SDK Adds TypeSafe Provider for Evaluation, Enhances Tool Discovery

The Vercel AI SDK has introduced a TypeSafe provider for experimental_evaluate, enabling native support for Choice, Score, and Boolean questions, structured JSON rubrics, and probability distributions.

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/llamaindex@3.0.103 · vercel/ai
Save

Vercel AI SDK has released updates across several packages, including @ai-sdk/llamaindex@3.0.103, @ai-sdk/klingai@4.0.44, @ai-sdk/huggingface@2.0.50, and @ai-sdk/hume@3.0.42. These updates introduce a new TypeSafe provider and enhance existing functionalities within the SDK.

The @ai-sdk/typesafe-ai@3.0.0 package now includes the TypeSafe provider for experimental_evaluate. This provider supports native Choice, Score, and Boolean questions within a single request. It also incorporates structured JSON rubrics, probability distributions, confidence metadata, and workflow serialization.

Key Points

  • The TypeSafe provider for experimental_evaluate is available in @ai-sdk/typesafe-ai@3.0.0.
  • This provider supports native Choice, Score, and Boolean questions in one request.
  • Structured JSON rubrics are now supported for evaluation.
  • Probability distributions and confidence metadata are included in the new provider.
  • Workflow serialization is a feature of the TypeSafe provider.
  • The ai@7.0.103 package adds mid-conversation tool discovery and updates for code-mode.
  • Node.js downloads are protected by DNS validation and connection pinning, even when frameworks wrap global fetch.

Context

According to the Vercel AI SDK Changelog, the updates were processed via a Changesets release GitHub action. The experimental_evaluate functionality aims to provide more robust and structured evaluation capabilities for AI models. The ai@7.0.103 update also addresses tool discovery during conversations, which can enhance the dynamic interaction capabilities of AI applications.

Why It Matters

The introduction of the TypeSafe provider for experimental_evaluate offers developers more precise and structured methods for evaluating AI model outputs. This can lead to more reliable assessments of model performance and behavior, particularly for tasks involving nuanced responses or complex scoring. Enhanced tool discovery during conversations can improve the adaptability of AI agents.

What To Do

  • Review the documentation for experimental_evaluate to understand the new TypeSafe provider capabilities.
  • Explore the use of structured JSON rubrics for evaluating model responses.
  • Test the mid-conversation tool discovery feature in ai@7.0.103 for code-mode applications.
  • Note the improved security for Node.js downloads related to DNS validation and connection pinning.

Keep Exploring

/techniques/ptcf /techniques/output-schema