test(vlm): fix always-true image assertion

[mickqian]

Summary

• make the person branch in the shared VLM single-image semantic check evaluate membership instead of a truthy string literal
• prevent test_vision_openai_server from accepting responses that miss all person/man/driver concepts

Validation

• python3 -m py_compile python/sglang/test/vlm_utils.py
• AST audit confirms there are no remaining bare truthy string operands in assert-or expressions under the VLM test utilities

---

CI States

Latest PR Test (Base): ⏳ Run #30347730447
Latest PR Test (Extra): ❌ Run #30347730327

[gemini-code-assist]

Code Review

This pull request fixes a logical bug in the assertion within verify_single_image_response where the string "person" was evaluated as a truthy value instead of checking its membership in text. The reviewer suggested improving the robustness of this check by making it case-insensitive.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

To make the assertion more robust against variations in model capitalization (e.g., if the model starts the response with 'Man' or 'Person'), consider performing a case-insensitive check by converting text to lowercase.

Suggested change

	"man" in text or "person" in text or "driver" in text
	"man" in text.lower() or "person" in text.lower() or "driver" in text.lower()

[mickqian]

/tag-and-rerun-ci