@ai-sdk/openai 2.0.143 adds prompt cache reuse for changed output schemas

The patch release of the OpenAI provider in Vercel's AI SDK adds support for positioned Responses API format updates. Changing a structured output schema can now keep earlier reasoning and reuse the prompt cache.

The release also adds a request-level text format override, which is marked as stable. It lets a single request set its text format without changing the wider configuration.

It exposes prompt cache comparison diagnostics on both generation and streaming results, so cache behavior can be checked from each call's output.

Agent builders who revise structured output schemas between steps may keep earlier reasoning and avoid losing cache reuse after the change. The per-request text format override makes it possible to vary output format for one call. The new diagnostics give a way to check whether cache reuse is actually happening in a given run.

The release is a patch, not a breaking change, so upgrading within the same major version should not require code changes.