{
  "consolidation_date": "2025-12-07",
  "statistics": {
    "total_issues": 27,
    "total_suggestions": 18,
    "total_missing_items": 16,
    "applied": 31,
    "modified": 19,
    "withdrawn": 1,
    "rejected": 10
  },
  "applied_items": [
    {
      "source_expert": "开发专家",
      "item_type": "issue",
      "item_index": 0,
      "severity": "high",
      "description": "外部数据源API访问可行性未验证",
      "status": "applied",
      "applied_content": "明确标注各数据源的授权方式，MVP阶段优先使用免费开放API（PubMed、bioRxiv），支持用户自带机构账号模式",
      "reason": "开发专家与产品经理达成共识，采用修改后方案"
    },
    {
      "source_expert": "开发专家",
      "item_type": "issue",
      "item_index": 2,
      "severity": "high",
      "description": "实时进度展示的技术实现方式不明确",
      "status": "applied",
      "applied_content": "在性能要求中增加进度反馈机制描述，包含预估完成时间和后台执行支持",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "开发专家",
      "item_type": "issue",
      "item_index": 4,
      "severity": "medium",
      "description": "'合理时间内完成'表述模糊",
      "status": "applied",
      "applied_content": "在验收标准中明确时间预期：简单问题30分钟内，复杂问题2小时内",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "开发专家",
      "item_type": "issue",
      "item_index": 7,
      "severity": "low",
      "description": "多数据源返回结果的格式标准化未考虑",
      "status": "applied",
      "applied_content": "在Agent能力定义中明确搜索Agent负责格式转换",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "开发专家",
      "item_type": "missing_item",
      "item_index": 2,
      "severity": "medium",
      "description": "缺少错误恢复机制说明",
      "status": "applied",
      "applied_content": "在性能要求中增加后台执行+完成通知支持",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "产品经理",
      "item_type": "issue",
      "item_index": 3,
      "severity": "medium",
      "description": "进度反馈机制描述不够具体",
      "status": "applied",
      "applied_content": "增加预估完成时间展示、后台执行+完成通知功能",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "产品经理",
      "item_type": "issue",
      "item_index": 5,
      "severity": "medium",
      "description": "边缘场景覆盖不足",
      "status": "applied",
      "applied_content": "在异常处理中增加用户问题模糊时的引导澄清机制",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "产品经理",
      "item_type": "issue",
      "item_index": 6,
      "severity": "medium",
      "description": "部分验收标准不够具体可测",
      "status": "applied",
      "applied_content": "在验收标准中明确复杂问题的定义和时间范围",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "产品经理",
      "item_type": "user_experience_concern",
      "item_index": 0,
      "severity": "medium",
      "description": "长时间等待的用户体验",
      "status": "applied",
      "applied_content": "增加预估完成时间、后台执行+完成通知功能",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "AI专家",
      "item_type": "issue",
      "item_index": 0,
      "severity": "high",
      "description": "引用准确性验收标准缺乏量化指标",
      "status": "applied",
      "applied_content": "分层定义验收指标：引用来源可追溯率=100%（刚性约束），引用格式准确率>=95%",
      "reason": "AI专家与开发专家达成共识"
    },
    {
      "source_expert": "AI专家",
      "item_type": "issue",
      "item_index": 4,
      "severity": "medium",
      "description": "'复杂问题处理'验收标准过于模糊",
      "status": "applied",
      "applied_content": "明确复杂问题的定义：涉及多种疾病类型、多种治疗方法的跨领域研究问题",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "AI专家",
      "item_type": "ai_risk",
      "item_index": 0,
      "severity": "high",
      "description": "引用幻觉风险",
      "status": "applied",
      "applied_content": "通过架构设计防范：报告生成Agent引用只能来自搜索返回列表，采用结构化输出+ID校验机制",
      "reason": "AI专家接受开发专家的简化方案"
    },
    {
      "source_expert": "AI专家",
      "item_type": "missing_item",
      "item_index": 2,
      "severity": "medium",
      "description": "未定义分析Agent处理单次任务的文献数量上限",
      "status": "applied",
      "applied_content": "采用分层处理策略：第一轮粗筛200篇，第二轮精读50-80篇核心文献",
      "reason": "AI专家接受开发专家的动态配置方案"
    },
    {
      "source_expert": "领域专家",
      "item_type": "issue",
      "item_index": 3,
      "severity": "medium",
      "description": "缺少专业术语规范化处理",
      "status": "applied",
      "applied_content": "建立精神科标准术语库（基于DSM-5/ICD-11），在问题解析阶段自动映射到标准术语",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "领域专家",
      "item_type": "issue",
      "item_index": 7,
      "severity": "low",
      "description": "输入示例可进一步优化",
      "status": "applied",
      "applied_content": "增加更专业的输入示例",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "领域专家",
      "item_type": "issue",
      "item_index": 8,
      "severity": "low",
      "description": "未明确处理预印本文献的风险提示",
      "status": "applied",
      "applied_content": "在数据源说明中增加预印本风险提示",
      "reason": "无争议，直接采纳"
    },
    {
      "source_expert": "领域专家",
      "item_type": "missing_item",
      "item_index": 0,
      "severity": "medium",
      "description": "缺少量表和评估工具识别功能",
      "status": "applied",
      "applied_content": "MVP阶段实现量表名称识别，Phase 2实现评分结果提取",
      "reason": "领域专家接受AI专家的分层实现建议"
    }
  ],
  "modified_items": [
    {
      "source_expert": "开发专家",
      "item_type": "suggestion",
      "item_index": 0,
      "severity": "medium",
      "original": "建议增加技术选型章节：明确开发语言（Python推荐）、Agent框架等核心技术决策",
      "modified": "建议在需求文档中明确技术约束（知识图谱存储能力、并行任务调度、实时进度反馈、医学术语标准化组件），具体技术选型留待技术设计阶段",
      "modifier": "产品经理+领域专家",
      "reason": "产品经理指出需求文档应保持技术中立，领域专家补充了医学术语标准化组件的必要性"
    },
    {
      "source_expert": "开发专家",
      "item_type": "issue",
      "item_index": 5,
      "severity": "medium",
      "original": "证据等级评估的实现复杂度被低估，建议结合文献元数据进行规则化判断",
      "modified": "采用分层策略：Phase 1做研究类型分类+规则化信息提取，Phase 2引入偏倚风险初筛；采用Cochrane偏倚风险评估工具框架，AI任务定位为信息提取而非判断",
      "modifier": "AI专家+领域专家",
      "reason": "AI专家和领域专家均指出完整GRADE评估超出AI能力边界，但研究类型分类是核心功能不可放弃"
    },
    {
      "source_expert": "开发专家",
      "item_type": "suggestion",
      "item_index": 2,
      "severity": "medium",
      "original": "建议细化MVP验收标准：增加定量指标如'引用准确率>=95%'、'去重准确率>=90%'",
      "modified": "分层定义验收指标：引用来源可追溯率=100%（刚性约束），引用格式准确率>=95%，去重准确率>=90%",
      "modifier": "AI专家",
      "reason": "AI专家提出将可追溯性与格式准确性分开定义更精准"
    },
    {
      "source_expert": "开发专家",
      "item_type": "tech_risk",
      "item_index": 1,
      "severity": "high",
      "original": "知识图谱去重准确性风险：分层去重方案",
      "modified": "分层去重：先DOI/PMID精确匹配，再UMLS/MeSH CUI映射实现跨语言术语对齐，最后标题相似度匹配；对无法匹配的实体采用保守策略",
      "modifier": "AI专家+领域专家",
      "reason": "两位专家均指出现有标准术语库可有效支持跨语言术语对齐"
    },
    {
      "source_expert": "产品经理",
      "item_type": "suggestion",
      "item_index": 2,
      "severity": "low",
      "original": "建议细化用户故事",
      "modified": "建议在验收标准部分增加基于用户视角的测试用例描述",
      "modifier": "开发专家",
      "reason": "开发专家指出当前场景描述已足够详细"
    },
    {
      "source_expert": "产品经理",
      "item_type": "issue",
      "item_index": 4,
      "severity": "medium",
      "original": "报告输出形式单一",
      "modified": "MVP阶段输出Markdown格式，Phase 2增加直接导出Word/PDF功能，英文报告作为后续版本考虑",
      "modifier": "开发专家+AI专家",
      "reason": "开发专家和AI专家均指出应分阶段实现"
    },
    {
      "source_expert": "产品经理",
      "item_type": "user_experience_concern",
      "item_index": 1,
      "severity": "medium",
      "original": "报告质量的可信度建立：显示证据强度评分",
      "modified": "显示研究类型分布（如包含3项RCT、5项队列研究等）替代AI直接评分，展示文献筛选逻辑",
      "modifier": "AI专家",
      "reason": "AI专家指出证据强度评分可能给用户造成虚假的专业感"
    },
    {
      "source_expert": "产品经理",
      "item_type": "issue",
      "item_index": 2,
      "severity": "high",
      "original": "缺少关键使用场景：医学生临床问题查证场景等",
      "modified": "MVP阶段聚焦学术研究场景，临床决策支持场景（诊断鉴别、治疗方案选择等）作为Phase 2扩展",
      "modifier": "领域专家",
      "reason": "领域专家指出临床决策支持与学术研究有本质区别，需差异化设计"
    },
    {
      "source_expert": "产品经理",
      "item_type": "user_experience_concern",
      "item_index": 2,
      "severity": "low",
      "original": "专业术语理解门槛：根据用户角色调整报告语言复杂度",
      "modified": "采用保持专业术语+增加解释注释的方式，不直接简化术语以避免损失专业精确性",
      "modifier": "领域专家",
      "reason": "领域专家指出精神科术语简化可能导致专业准确性损失"
    },
    {
      "source_expert": "AI专家",
      "item_type": "suggestion",
      "item_index": 0,
      "severity": "medium",
      "original": "增加引用验证Agent角色",
      "modified": "通过架构设计防范引用幻觉：结构化输出+ID校验，作为报告生成Agent的内置功能",
      "modifier": "开发专家",
      "reason": "开发专家指出独立Agent过度设计，架构约束即可解决"
    },
    {
      "source_expert": "AI专家",
      "item_type": "suggestion",
      "item_index": 3,
      "severity": "medium",
      "original": "MVP阶段先实现用户对搜索策略的确认功能",
      "modified": "提供搜索策略确认的可选功能：默认自动执行，高级用户可开启策略预览，结果页面提供调整后重新生成入口",
      "modifier": "产品经理",
      "reason": "产品经理指出强制确认会打断用户流程，采用可选模式"
    },
    {
      "source_expert": "AI专家",
      "item_type": "suggestion",
      "item_index": 5,
      "severity": "low",
      "original": "明确定义单次任务的文献处理上限（如50篇）",
      "modified": "采用分层处理策略（粗筛200篇+精读50-80篇）+动态配置，而非固定数值",
      "modifier": "开发专家+产品经理",
      "reason": "开发专家指出50篇过于保守，产品经理建议用户可选范围"
    },
    {
      "source_expert": "AI专家",
      "item_type": "issue",
      "item_index": 2,
      "severity": "high",
      "original": "知识图谱的实体语义去重能力要求过高",
      "modified": "跨语言术语对齐可通过整合现有标准术语库（ICD-11、UMLS、MeSH）实现，技术难度低于从零构建",
      "modifier": "领域专家",
      "reason": "领域专家指出现有医学术语标准化资源成熟度被低估"
    },
    {
      "source_expert": "AI专家",
      "item_type": "ai_risk",
      "item_index": 1,
      "severity": "high",
      "original": "证据等级评估不可靠风险，建议MVP阶段简化为研究类型分类",
      "modified": "采用Cochrane偏倚风险评估工具框架，AI任务定位为从文献中提取信息填充清单，基于清单结果进行规则化判断",
      "modifier": "领域专家",
      "reason": "领域专家提供了现有评估框架可转化为结构化检查清单的方案"
    },
    {
      "source_expert": "领域专家",
      "item_type": "issue",
      "item_index": 0,
      "severity": "high",
      "original": "缺少诊断标准版本标注功能，自动识别并标注每篇文献的诊断标准版本",
      "modified": "MVP阶段增加诊断标准注意事项提醒章节+关键词识别，Phase 2尝试自动识别",
      "modifier": "开发专家+产品经理",
      "reason": "开发专家指出自动识别技术实现有挑战，应分阶段实现"
    },
    {
      "source_expert": "领域专家",
      "item_type": "issue",
      "item_index": 1,
      "severity": "high",
      "original": "证据等级评估方法未明确，建议建立自动识别和分级逻辑",
      "modified": "MVP阶段实现研究类型分类，Phase 2尝试偏倚风险初筛，完整GRADE评估定位为人工任务",
      "modifier": "AI专家",
      "reason": "AI专家指出GRADE完整评估超出AI可靠能力边界"
    },
    {
      "source_expert": "领域专家",
      "item_type": "issue",
      "item_index": 2,
      "severity": "high",
      "original": "未涵盖临床试验注册库",
      "modified": "ClinicalTrials.gov列入Phase 2扩展数据源，WHO ICTRP根据技术条件评估后决定",
      "modifier": "开发专家+产品经理",
      "reason": "开发专家指出WHO ICTRP缺乏稳定公开API，产品经理建议控制MVP范围"
    },
    {
      "source_expert": "领域专家",
      "item_type": "suggestion",
      "item_index": 7,
      "severity": "medium",
      "original": "MVP阶段也应包含诊断标准版本标注和基本的证据等级评估",
      "modified": "MVP阶段实现诊断标准关键词识别+研究类型分类，标注为AI初步分类，完整循证医学评估功能在Phase 2完善",
      "modifier": "开发专家+产品经理+AI专家",
      "reason": "多位专家指出原建议对MVP范围定义过于激进"
    },
    {
      "source_expert": "AI专家",
      "item_type": "suggestion",
      "item_index": 1,
      "severity": "high",
      "original": "将证据等级评估任务降级为研究类型分类",
      "modified": "保留证据等级评估功能，采用结构化规则评估方式：基于研究类型、样本量、盲法等客观元数据进行规则化判断，标注AI局限性",
      "modifier": "开发专家+产品经理+领域专家",
      "reason": "三位专家均指出完全降级会损害产品核心价值"
    }
  ],
  "rejected_items": [
    {
      "source_expert": "开发专家",
      "item_type": "issue",
      "item_index": 1,
      "severity": "high",
      "description": "知识图谱技术选型和存储方案未明确",
      "status": "rejected",
      "reason": "属于技术设计阶段内容，需求文档应保持技术中立"
    },
    {
      "source_expert": "开发专家",
      "item_type": "issue",
      "item_index": 3,
      "severity": "medium",
      "description": "Agent通信机制未定义",
      "status": "rejected",
      "reason": "属于技术设计阶段内容，需求文档应保持技术中立"
    },
    {
      "source_expert": "开发专家",
      "item_type": "missing_item",
      "item_index": 0,
      "severity": "medium",
      "description": "缺少技术栈选型说明",
      "status": "rejected",
      "reason": "属于技术设计阶段内容"
    },
    {
      "source_expert": "开发专家",
      "item_type": "missing_item",
      "item_index": 1,
      "severity": "medium",
      "description": "缺少LLM模型选型和调用方式",
      "status": "rejected",
      "reason": "属于技术设计阶段内容"
    },
    {
      "source_expert": "开发专家",
      "item_type": "missing_item",
      "item_index": 5,
      "severity": "low",
      "description": "缺少监控和日志方案",
      "status": "rejected",
      "reason": "属于技术设计阶段内容"
    },
    {
      "source_expert": "产品经理",
      "item_type": "missing_item",
      "item_index": 0,
      "severity": "low",
      "description": "竞品分析",
      "status": "rejected",
      "reason": "属于产品规划阶段内容，非需求文档范畴"
    },
    {
      "source_expert": "产品经理",
      "item_type": "missing_item",
      "item_index": 1,
      "severity": "low",
      "description": "用户旅程地图",
      "status": "rejected",
      "reason": "当前场景描述已足够详细"
    },
    {
      "source_expert": "产品经理",
      "item_type": "missing_item",
      "item_index": 5,
      "severity": "low",
      "description": "数据导出与集成（Zotero、EndNote等）",
      "status": "rejected",
      "reason": "超出用户原始需求范围，可作为后续版本考虑"
    },
    {
      "source_expert": "领域专家",
      "item_type": "issue",
      "item_index": 4,
      "severity": "medium",
      "description": "未区分药物治疗与非药物治疗的文献分析逻辑",
      "status": "rejected",
      "reason": "过于细化的领域逻辑，可在技术设计阶段考虑"
    },
    {
      "source_expert": "领域专家",
      "item_type": "issue",
      "item_index": 6,
      "severity": "medium",
      "description": "未提及临床实践指南的整合",
      "status": "rejected",
      "reason": "超出用户原始需求范围，指南整合复杂度高，可作为后续版本考虑"
    }
  ],
  "withdrawn_items": [
    {
      "source_expert": "产品经理",
      "item_type": "missing_item",
      "item_index": 4,
      "description": "移动端适配",
      "withdrawn_by": "产品经理",
      "reason": "开发专家指出目标用户的核心使用场景是桌面端长时间研究工作，移动端投入产出比低"
    }
  ],
  "conflict_resolutions": [
    {
      "topic": "证据等级评估功能范围",
      "conflicting_parties": ["AI专家（建议降级为研究类型分类）", "领域专家（坚持证据等级是核心价值）", "产品经理（强调用户核心需求）"],
      "resolution": "保留证据等级评估功能，但采用结构化规则评估方式降低AI风险",
      "reason": "按用户价值优先原则裁决，证据等级评估是用户明确需求，但采纳AI专家关于风险控制的建议"
    },
    {
      "topic": "MVP阶段专业功能范围",
      "conflicting_parties": ["领域专家（要求诊断标准自动识别+完整证据评估）", "开发专家+产品经理（控制MVP范围）"],
      "resolution": "MVP阶段实现诊断标准关键词识别+研究类型分类，完整功能留待Phase 2",
      "reason": "按技术可行性优先原则裁决，同时保留核心专业价值"
    },
    {
      "topic": "引用幻觉防范机制",
      "conflicting_parties": ["AI专家（建议增加引用验证Agent）", "开发专家（架构设计即可解决）"],
      "resolution": "采用开发专家方案：结构化输出+ID校验作为报告生成Agent内置功能",
      "reason": "按技术可行性优先原则裁决，更简洁的架构方案同样有效"
    },
    {
      "topic": "文献处理数量上限",
      "conflicting_parties": ["AI专家（建议50篇固定上限）", "开发专家+产品经理（动态配置+分层处理）"],
      "resolution": "采用分层处理策略：粗筛200篇+精读50-80篇，上限动态可配置",
      "reason": "按技术可行性优先原则裁决，分层策略更符合实际工程需求"
    }
  ],
  "key_improvements_summary": [
    "明确了外部数据源的授权方式和优先级，MVP阶段聚焦免费开放API数据源",
    "增加了术语规范化处理能力（基于DSM-5/ICD-11/MeSH/UMLS）",
    "细化了验收标准：引用来源可追溯率100%（刚性约束），引用格式准确率>=95%，去重准确率>=90%",
    "建立了分层文献处理策略（粗筛200篇+精读50-80篇）",
    "增加了引用幻觉防范机制（结构化输出+ID校验）",
    "明确了证据等级评估的分阶段实现路径（MVP研究类型分类，Phase 2偏倚风险初筛）",
    "增加了进度反馈机制的具体描述（预估完成时间、后台执行+通知）",
    "增加了精神科专业功能（量表名称识别、诊断标准关键词识别、方法学注意事项章节）",
    "增加了预印本来源的风险提示",
    "明确了报告透明性要求（研究类型分布、文献筛选逻辑、AI局限性说明）"
  ]
}