fix/feat: P0 bug fixes, type annotations, docstrings, and more tests

[MisterBrookT]

Summary

• P0 — 致命 bug 修复

  • prompts/gen_text2image.py: 修复错误的 nanochart 导入 → igenbench
  • eval_engine.py: call_image_understanding 返回值可能是 str，增加 isinstance(dict) 判断，避免 AttributeError
  • vis_item.py: generation 字段类型由错误的 Optional[dict] 改为 Dict[str, Any]
  • utils/io.py: 函数名拼写错误 senmantic → semantic，并添加 separator 不存在时的 ValueError

• P1 — 元数据与类型

  • pyproject.toml 补全 authors、keywords、classifiers、[project.urls]
  • 所有 CLI 命令函数添加 -> None 返回类型；GenEngine.text2image 添加 -> PILImage
  • caller_registry.py / llm_caller.py / client.py 补全 docstring 和返回类型注解
  • CI 添加 pytest --cov 和 coverage artifact 上传

• 新增测试（41 个，较之前 17 个增加 141%）

  • test_io.py：9 个测试覆盖 extract_from_markdown 和 split_semantic_and_data_in_t2i_prompt
  • test_eval_engine.py：6 个测试覆盖 EvalEngine（mock LLMClient，无需 API key）
  • test_caller_registry.py：4 个测试覆盖 @register_caller 装饰器
  • test_cli_score.py：5 个测试覆盖 igenbench score CLI 命令

• P2 — 文档

  • AGENTS.md：补全废弃 Question → EvalEntry 的字段迁移说明

Test plan

• uv run pytest tests/ -v — 41/41 passed 本地验证
• uv run ruff check . — All checks passed
• uv run ruff format --check . — All files formatted
• CI 三个 Python 版本全绿

---

Generated by Claude Code

[gemini-code-assist]

Code Review

This pull request adds comprehensive type annotations, docstrings, metadata, and unit tests across the repository, alongside minor bug fixes such as correcting an import path and handling non-dict LLM responses. Reviewer feedback highlights several opportunities to improve robustness through defensive programming, specifically by handling potential None or empty values in LLM responses and prompt inputs to prevent runtime errors like AttributeError and TypeError.

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]

If the LLM provider returns an empty response or None (for example, due to safety filters or generation limits), passing it directly to extract_from_markdown will raise an AttributeError when calling .strip(). Adding a check to return an empty string when text_response is falsy prevents runtime crashes.

        text_response = self._caller.generate_text(model, prompt, **kwargs)
        if not text_response:
            return ""
        return extract_from_markdown(text_response)

[gemini-code-assist]

Similar to call_text_generation, if understand_image returns None or an empty response, passing it directly to extract_from_markdown will cause an AttributeError. Adding a defensive check here ensures the engine handles empty or blocked responses gracefully.

Suggested change

	text_response = self._caller.understand_image(
	model, prompt, image_path, **kwargs
	)
	return extract_from_markdown(text_response)
	text_response = self._caller.understand_image(
	model, prompt, image_path, **kwargs
	)
	if not text_response:
	return ""
	return extract_from_markdown(text_response)

[gemini-code-assist]

If response is None (e.g., if the API call failed or returned an empty response), calling str(response) will set analysis to the literal string "None". It is better to default to an empty string "" to avoid saving "None" as the analysis result.

Suggested change

	if isinstance(response, dict):
	analysis = response.get("analysis", "")
	answer = response.get("answer", "")
	else:
	analysis = str(response)
	answer = ""
	if isinstance(response, dict):
	analysis = response.get("analysis", "")
	answer = response.get("answer", "")
	else:
	analysis = str(response) if response is not None else ""
	answer = ""

[gemini-code-assist]

If t2i_prompt is None or empty, checking separator not in t2i_prompt will raise a TypeError. Adding a defensive check at the beginning of the function ensures we raise a clear, descriptive ValueError instead of a generic runtime error.

    if not t2i_prompt:
        raise ValueError("Prompt cannot be empty or None.")
    separator = "The given data is:"
    if separator not in t2i_prompt: