The Vercel AI SDK has released updates across several packages, including @ai-sdk/workflow@2.0.4 and @ai-sdk/workflow-harness@1.0.82. These updates introduce new features and dependency changes, with a focus on enhanced model integration and video processing capabilities.
Key Points
- The
@ai-sdk/deepseek@3.0.30package now supports DeepSeek V4 Flash Vision Exp for image input. - The
@ai-sdk/deepseek@3.0.30package also includes support for the Files API. - The
@ai-sdk/gateway@4.0.60package implements ahandleWebhookOptionon its video model. - This webhook implementation allows
generateVideo({ webhook })to register a factory URL as the gateway'scallbackUrl. - The webhook functionality enables awaiting video delivery instead of relying on polling.
Context
According to the Vercel AI SDK Changelog, these updates were initiated by a Changesets release GitHub action. The process is designed to automatically publish packages to npm upon merging the relevant pull request. The updates also include various patch changes and updated dependencies across packages like ai@7.0.74, @ai-sdk/angular@3.0.74, @ai-sdk/azure@4.0.47, @ai-sdk/code-mode@1.0.31, @ai-sdk/harness@1.0.82, and @ai-sdk/harness-acp@1.0.20.
Why It Matters
These updates enhance the Vercel AI SDK's capabilities by integrating a new vision model and improving asynchronous video processing. Builders can leverage these features to develop applications that require advanced image input and more efficient video generation workflows.
What To Do
- Review the updated
@ai-sdk/deepseekdocumentation for details on DeepSeek V4 Flash Vision Exp image input and Files API usage. - Explore the
handleWebhookOptionimplementation in@ai-sdk/gatewayfor video model integrations. - Test the
generateVideo({ webhook })functionality to understand its impact on delivery mechanisms. - Note the updated dependency versions for
aiand other related@ai-sdkpackages in your projects.
