ai@7.0.153
Vercel AI SDK 7.0.153 fixes structured output parsing in non-streaming Responses API when multiple assistant messages contain JSON, ensuring OpenAI and Azure Responses models select the final answer properly.
Why this signal matters
The ai@7.0.153 patch release addresses an issue in non-streaming Responses API structured output parsing when multiple assistant messages contain JSON. Providers are now able to supply separate structured-output text for parsing. This update enables OpenAI and Azure Responses models to select the last final answer while preserving all generated content, annotations, conversation history, and token usage metrics. Dependency updates include @ai-sdk/provider@4.0.34, @ai-sdk/provider-utils@5.0.68, and @ai-sdk/gateway@4.0.126.
Actionable summary
ai@7.0.153 resolves non-streaming Responses API structured output parsing failures triggered by multiple assistant messages containing JSON. It permits providers to supply distinct structured-output text, ensuring OpenAI and Azure Responses models target the last final answer without losing generated content, annotations, token usage, or conversation history.
- Agent usefulness
- 77/100
- Confidence
- 96%
- Canonical data
- JSON + Markdown
What builders should check
- Upgrade ai to version 7.0.153 if using non-streaming structured outputs with OpenAI or Azure Responses models.