Vercel’s AI SDK 3.1 Unleashes Seamless AI Integration

Recently, Vercel, the frontend cloud provider, has taken over the AI integration tool, ModelFusion. The benefit of this acquisition is available in the latest release of Vercel’s AI SDK v.3.1. The recent release and takeover bring Vercel closer to offering a complete TypeScript framework for building AI applications. According to team members, ModelFusion would have the capability to streamline the process of integrating AI models for developers. In fact, the core of Vercel’s AI SDK contains functionality that allows direct communication with large language models (LLMs). Consequently, developers can now generate LLMs from various sources and select an LLM and source that best suits their requirements.

The latest update is split into three main categories:

• The first part is the AI SDK Core. This packs an API capable of creating structured objects, text, and tool calls using large language models (LLMs). According to Vercel’s Vice President of Product, AI, Jared Palmer, and Software Engineer Lars Grammel, they drew influences from initiatives like Prisma and Drizzle to shape the AI SDK Core similar to ORM-style abstraction for LLMs. These fresh APIs offer unified, raw-level elements to help interact with LLMs across any JavaScript environment and sidestep the quirks of major model providers. Furthermore, integrating LLMs has been made easier by reducing it to two simple questions, one on the type of data text or structured object you wish to create and another on whether you prefer your data delivered incrementally or streaming all-at-once.

• The second component is the AI SDK UI. This part features a set of framework-agnostic hooks tailored for quickly building chat interfaces. This simplifies AI interfacing by introducing three main hooks: useChat, useCompletion, and useAssistant. Developers can write less than 50 lines of code using the streamText function.

• The third part is AI SDK RSC: Data is available in realtime for AI applications built by developers using the AI SDK RSC. This library streams generative user interfaces with React Server Components (RSC).

Visit their site to check out examples and experiment with different models using the AI playground.