mirror of
https://github.com/docker/docker-agent.git
synced 2026-11-04 18:30:09 +00:00
Omit thinkingConfig for title-generation requests so image-capable Gemini models can return plain text titles. Increase the shared title output budget from 20 to 128 tokens to accommodate hidden reasoning while keeping the visible title separately capped at 50 characters. Preserve explicit no-thinking behavior for MCP sampling and ordinary requests, including Gemini 3 minimum reasoning settings and Gemini 2.5 zero-budget suppression. Keep user-configured thinking budgets outside title generation unchanged and test both paths.