文件内容
evals/evals.json
{
"skill_name": "pdf-finance-parser",
"evals": [
{
"id": 1,
"prompt": "解析这份PDF:samples/documents/sample_001.pdf,输出到 samples/outputs/sample_001",
"expected_output": "完成端到端解析并落盘 output.md + output.json",
"expected_exit_code": 0,
"expected_stdout_contains": ["task_id", "COMPLETED", "page_count"],
"files": ["samples/documents/sample_001.pdf"]
},
{
"id": 2,
"prompt": "用 detail 模式精细解析 samples/documents/complex_table.pdf",
"expected_output": "parse_mode=detail,cells 数值规整字段填充率应高于 normal 模式",
"expected_exit_code": 0,
"expected_cmd_contains": ["--parse-mode", "detail"],
"files": ["samples/documents/complex_table.pdf"]
},
{
"id": 3,
"prompt": "只解析这份长报告的前10页:samples/documents/long_report.pdf",
"expected_output": "page_count=10,total_pages>10,wall_time 与页数成正比",
"expected_exit_code": 0,
"expected_cmd_contains": ["--pages", "1-10"],
"files": ["samples/documents/long_report.pdf"]
},
{
"id": 4,
"prompt": "异步提交解析任务:samples/documents/sample_002.pdf",
"expected_output": "submit 返回 task_id 与 eta;check-and-notify 完成实际解析",
"expected_exit_code": 0,
"expected_cmd_contains": ["submit", "check-and-notify"],
"files": ["samples/documents/sample_002.pdf"]
},
{
"id": 5,
"prompt": "解析这张扫描件图片:samples/documents/scanned_invoice.png",
"expected_output": "自动识别为图片输入,单页 VLM 调用完成解析",
"expected_exit_code": 0,
"expected_stdout_contains": ["page_count"],
"files": ["samples/documents/scanned_invoice.png"]
}
]
}