July 2026 performance summary: Reliability and onboarding gains across vercel/ai and vercel/ai-chatbot. Key outcomes include dynamic chat transport refresh to prevent stale state, robust tool input handling for Bedrock/Anthropic, evergreen documentation for use-ai-sdk, user authentication foundation, and lint/CI improvements—resulting in stronger business value and faster delivery.

June 2026 monthly summary for vercel/ai focused on expanding model support, improving reliability, and strengthening developer UX. Delivered core model provider enhancements, enabling structured outputs and expanded model typing for production-grade usage. Implemented routing for tuned Google Vertex models and added scribe_v2 support to ElevenLabs, reducing model-not-found and misrouting errors. Enabled structured outputs for kimi-k2.6 and kimi-k2.7-code, and refreshed the Kimi model list to the latest IDs to ensure compatibility with current deployments. Streamlined the codebase by removing an outdated skills symlink, aligning with native support for .agents/skills. Enhanced user guidance and workflow UX through documentation updates (version mismatch warnings, onEnd usage, token guidance) to reduce runtime toil. Overall, these changes improve model coverage, reliability, and developer experience while unlocking faster iteration and safer deployments for production workloads.