Vercel AI SDK xAI Adapter 3.0.133 Patch Explained

The xAI provider adapter @ai-sdk/xai for Vercel AI SDK has released version 3.0.133. This is a patch-level update focused on fixing a schema handling issue in tool calling scenarios. While the version bump is minor, for developers relying on xAI (Grok series models) for structured output and function calling, this kind of fix can directly impact application stability.

What Changed: Tool Schema Handling Fix

This update contains a single change from commit 8825c65:

fix(xai): preserve additionalProperties: false in tool schemas

In short, this patch ensures that when passing tool definitions to xAI models, the additionalProperties: false field in JSON Schema is correctly preserved.

additionalProperties: false is a critical JSON Schema constraint that explicitly tells the model: only properties declared in the schema are allowed in the object — no extra fields permitted. In structured output and function calling scenarios, this constraint is essential for ensuring strict model output. If accidentally dropped, the model may return responses with extra fields, causing downstream parsing failures or type validation errors.

From a technical standpoint, additionalProperties: false typically gets lost during schema transformation or serialization within the SDK. Different model providers have subtle differences in their tool schema format requirements, so the adapter layer needs to map and clean up the original schema. During this process, if the code only performs a shallow copy of properties, or uses a serialization library that filters out "unknown fields," keywords like additionalProperties — which are non-standard but valid JSON Schema keywords — can be silently dropped. OpenAI's function calling spec and xAI may also handle this field differently, and a slight misalignment when reconciling the two can easily produce this kind of omission.

Why This Fix Matters

Tool calling is one of the core capabilities in modern AI application development. Developers define structured tool schemas so that models return parameters in a predictable format, which code then executes. If the SDK loses additionalProperties: false during schema transformation, the output constraints on the model are loosened, potentially introducing edge-case bugs that are hard to debug.

For projects using xAI Grok models to build Agents, structured data extraction, or API orchestration, the correctness of these low-level adapters directly determines the reliability of business logic. Small as this patch is, it's a necessary piece to ensure that "strict mode" works as expected.

---

Here's how tool calling (Tool Calling / Function Calling) works: when calling a model API, developers attach a set of "tool definitions" to the request. Each tool includes a name, description, and a JSON Schema for its parameters. When generating a response, if the model determines it needs to use a tool, it outputs a structured parameter object matching that schema — rather than plain text. The application code then takes those parameters, calls the corresponding local function or external API, and feeds the result back to the model, completing one "think–act–observe" cycle. This is the foundational mechanism for building AI Agents, and it's why schema constraint correctness is so critical — if the model's output parameters deviate from the expected format, the entire tool-calling pipeline can break down.

About Vercel AI SDK

Vercel AI SDK is one of the most popular TypeScript AI development toolkits available today, with over 26,000 stars and 5,100 forks on GitHub. Through a unified interface abstraction, it lets developers connect to multiple model providers — including OpenAI, Anthropic, and xAI — in a consistent way. @ai-sdk/xai is the official adapter package for xAI within this ecosystem.

This "provider adapter" architecture means that each model vendor's API quirks — such as schema format differences, parameter naming conventions, and tool calling protocols — are encapsulated in separate packages. As a result, patch updates targeting individual providers tend to be fairly frequent. Developers are encouraged to keep their dependencies up to date to benefit from compatibility and stability fixes.

Upgrade Recommendation

If your project uses xAI models and involves tool calling or structured output, it's recommended to upgrade @ai-sdk/xai to 3.0.133. The upgrade is straightforward:

npm install @ai-sdk/xai@3.0.133

As a patch release, this update contains no breaking changes and is safe to apply. If you've previously encountered issues with Grok models returning extra fields or failing structured output validation, this fix is likely the solution you've been waiting for.