Vercel has released updates to its AI SDK, including version @ai-sdk/vercel@3.0.24, @ai-sdk/vue@4.0.54, @ai-sdk/voyage@2.0.22, and @ai-sdk/valibot@3.0.22. These updates introduce a new audio provider and refine how language model instructions and agent callbacks are managed.
Key Points
- The @ai-sdk/fish-audio@3.0.0 package adds a new Fish Audio provider, supporting speech and transcription models.
- ai@7.0.54 allows ToolLoopAgent
prepareCallcallbacks to read and override the top-level reasoning option. - A
defaultInstructionsMiddlewarehas been added for applying default language model instructions while preserving call-level overrides. - Preceding assistant messages are now preserved when regenerating a response.
- @ai-sdk/anthropic@4.0.32 rejects spliced Anthropic generations but permits duplicate message start events for the active message.
Context
According to the Vercel AI SDK changelog, these updates are part of ongoing development to enhance the SDK's capabilities and flexibility. The changes were released via a GitHub action, with packages published to npm automatically.
Why It Matters
These updates provide developers with more control over language model interactions and agent behavior within the Vercel AI SDK. The addition of a new audio provider expands the types of applications that can be built using the SDK, while improved instruction handling and message preservation can lead to more consistent and robust AI-powered experiences.
What To Do
- Review the @ai-sdk/fish-audio@3.0.0 documentation to explore its speech and transcription model capabilities.
- Examine the ToolLoopAgent
prepareCallcallbacks to understand how to read and override reasoning options. - Investigate the
defaultInstructionsMiddlewarefor managing language model instructions. - Note the changes in @ai-sdk/anthropic@4.0.32 regarding spliced generations and message start events.
