Vercel has released updates to its AI SDK, addressing telemetry attribution and enhancing integration with Amazon Bedrock. The updates, detailed across several package releases, include version ai@7.0.44 and @ai-sdk/amazon-bedrock@5.0.38.
Key Points
- The ai@7.0.44 package includes a fix for telemetry attribution for language model calls that resolve to a different response model.
- The @ai-sdk/amazon-bedrock@5.0.38 package introduces support for video inputs within Converse messages.
- Multiple related packages, such as @ai-sdk/angular@3.0.44, @ai-sdk/code-mode@1.0.1, and @ai-sdk/harness@1.0.50, have updated dependencies on ai@7.0.44.
- Other updated packages include @ai-sdk/harness-claude-code@1.0.51, @ai-sdk/harness-codex@1.0.52, @ai-sdk/harness-deepagents@1.0.49, @ai-sdk/harness-opencode@1.0.51, @ai-sdk/harness-pi@1.0.50, and @ai-sdk/langchain@3.0.44.
Context
According to the Vercel AI SDK Changelog, these updates are part of routine releases managed by a GitHub action. The changes are published to npm automatically upon merging the relevant pull requests.
Why It Matters
These updates provide developers using the Vercel AI SDK with improved telemetry accuracy and expanded multimodal capabilities through Amazon Bedrock, potentially enabling more sophisticated applications that handle video input.
What To Do
- Review the updated @ai-sdk/amazon-bedrock documentation for details on implementing video inputs in Converse messages.
- Check the changelog for ai@7.0.44 to understand the specifics of the telemetry attribution fix.
- Update your project dependencies to the latest versions of the Vercel AI SDK packages to incorporate these changes.
- Test existing language model call implementations to ensure telemetry is accurately reported after updating.