v0.8.1

Bug Fixes

• Completions: preserve streamed reasoning_details wire order and opaque encrypted entries (#118).
• Google: use the correct embedding batch limit and fully paginate generative and embedding model discovery (#120).

Notes

• The chat-model filtering change from #119 was fully reverted by #121, so it has no net behavior change in this release.

View changes on GitHub