内部知识库 60%] + TS[tavily_search
外部搜索 40%] + AI[analyze_image
图片分析] + end + + subgraph Medical["医疗诊断工具"] + COD[chain_of_diagnosis
5步诊断推理链] + CONF[evaluate_diagnosis_confidence
置信度评估] + IMG[search_pathology_images
病理图片搜索] + GUIDE[generate_medical_guide
就医指南] + end + + subgraph Game["诊断模拟游戏"] + START[start_diagnosis_game] + ACTION[diagnosis_action] + end + + subgraph Visual["可视化工具 9个"] + VIS[knowledge_graph / diagnosis_flow / medical_chart
radar_chart / timeline / gantt_chart
quadrant_chart / state_diagram / sankey_diagram] + end + + Q --> A --> B + B -->|医学问答| KB + B -->|医学问答| TS + B -->|图片分析| AI + B -->|诊断分析| COD + B -->|诊断游戏| START + B -->|可视化| VIS + B -->|就医咨询| GUIDE + + KB --> R[生成回答] + TS --> R + AI --> R + COD --> CONF --> R + START --> ACTION --> R + VIS --> R + IMG --> R + GUIDE --> R +``` + +### 5.3 数据流向 + +```mermaid +flowchart LR + subgraph Input["输入"] + I1[文本问题] + I2[病理图片] + I3[按钮点击] + end + + subgraph Process["处理"] + Agent[AI助手] + Tools[MCP工具] + KB[(知识库 60%)] + Web((互联网 40%)) + end + + subgraph Output["输出"] + O1[Markdown文本] + O2[Mermaid图表] + O3[病理图片] + O4[交互按钮] + end + + I1 --> Agent + I2 --> Agent + I3 --> Agent + Agent --> Tools + Tools --> KB + Tools --> Web + KB --> O1 + Web --> O1 + Tools --> O2 + Tools --> O3 + Tools --> O4 +``` + +--- + +## 六、前端改进 + +### 6.1 新增组件 + +| 组件 | 文件 | 功能 | +|------|------|------| +| 病理图片画廊 | `PathologyImageGallery.tsx` | 展示和预览病理图片 | +| 置信度卡片 | `DiagnosisConfidenceCard.tsx` | 显示诊断置信度和风险等级 | +| 来源标签 | `SourceTag.tsx` | 标注信息来源 [内部]/[外部] | + +### 6.2 修改组件 + +| 组件 | 修改内容 | +|------|----------| +| `markdownRenderer.tsx` | 新增 `[btn:xxx]` 按钮解析和渲染 | +| `chatLeftSidebar.tsx` | 新增"清空所有对话"功能 | +| `conversationService.ts` | 新增 `deleteAll` 批量删除方法 | +| `MedicalVisualizationPanel.tsx` | 去除硬编码，支持通用病理学 | + +--- + +## 部署说明 + +### 8.1 环境要求 + +- Nexent v1.7.7+ +- Docker & Docker Compose +- Python 3.10+ +- Node.js 18+ + +### 8.2 部署步骤 + +1. **部署 Nexent 平台** + ```bash + docker-compose up -d + ``` + +2. **导入 Agent 配置** + ```bash + docker exec -i nexent-postgres psql -U postgres -d nexent < update_prompt_btn.sql + ``` + +3. **配置模型** + - 在 Nexent 管理界面配置各模型 API + +4. **启动使用** + - 访问 http://localhost:3000 + - 选择"病理学AI助手"开始对话 + +--- + +## 九、使用示例 + +### 9.1 医学问答 + +``` +用户: HIV感染的诊断标准是什么？ +AI: [调用 knowledge_base_search + tavily_search] + [融合内部60% + 外部40%结果] + 返回诊断标准说明... +``` + +### 9.2 诊断推理 + +``` +用户: 患者发热、淋巴结肿大、皮疹，请分析 +AI: [调用 chain_of_diagnosis] + Step1: 症状分析 - 三联征提示... + Step2: 病史关联 - ... + Step3: 鉴别诊断 - ... + Step4: 检查建议 - ... + Step5: 初步结论 - ... + [调用 evaluate_diagnosis_confidence] + 置信度: 85% | 风险等级: MEDIUM +``` + +### 9.3 诊断游戏 + +``` +用户: 开始诊断游戏 +AI: [调用 start_diagnosis_game] + + 🎮 诊断模拟游戏 + 难度: 中级 | 病例类型: 感染性疾病 + + 患者信息: 男性，35岁，主诉发热3天... + + [btn:询问发热情况] [btn:询问其他症状] [btn:询问既往史] +``` + +--- + +## 📝 应用价值 + +### 医学教育 + +- 医学生病理学习辅助 +- 住院医师规范化培训 +- 在线医学教育平台 +- 临床思维训练（诊断游戏） + +### 临床辅助 + +- 病理医生诊断辅助决策 +- 基层医院诊断能力提升 +- 远程病理会诊支持 +- 结构化诊断报告生成 + +### 科研应用 + +- 病理知识图谱构建 +- 医学AI研究数据积累 +- 诊断算法优化验证 +- 可视化数据展示 + +### 社会价值 + +- 缓解病理医生严重短缺（缺口3万+） +- 提升诊断标准化和准确性 +- 推动优质医疗资源下沉 +- 降低误诊漏诊风险 + +--- + +## ⚠️ 安全声明 + +**重要提示**： +- 本AI助手仅供学习和参考使用 +- 不能替代专业医生的诊断和治疗建议 +- 如有健康问题，请及时就医 +- 所有诊断建议均需专业医师确认 + +--- + +## 📄 许可证 + +Apache License 2.0 + +--- + +## 👥 贡献者 + +**量子工坊团队** + +开放原子基金会 ModeEngine AI 创新应用学习赛 diff --git a/pathology-ai/agent-config.md b/pathology-ai/agent-config.md new file mode 100644 index 000000000..739e45d4d --- /dev/null +++ b/pathology-ai/agent-config.md @@ -0,0 +1,80 @@ +# 🤖 智能体配置说明 + +## 智能体基本信息 + +| 配置项 | 值 | +|--------|-----| +| **Agent ID** | 13 | +| **名称** | 病理学AI助手 | +| **类型** | 医疗诊断辅助智能体 | +| **基础模型** | GPT-4 / 兼容OpenAI API | +| **max_steps** | 25 | + +## 业务描述 + +病理学AI助手具备以下能力： + +1. **医学知识问答** - 回答病理学、临床医学问题 +2. **诊断推理** - Chain-of-Diagnosis结构化诊断 +3. **置信度评估** - 评估回答可靠性和风险等级 +4. **交互式诊断练习** - 模拟游戏训练临床思维 +5. **医学可视化** - 知识图谱、流程图生成 + +## 工具配置 + +| 工具名称 | 来源 | 功能 | +|----------|------|------| +| `knowledge_base_search` | 内置 | 本地知识库搜索 | +| `tavily_search` | 内置 | 外部互联网搜索 | +| `analyze_image` | 内置 | 图片分析 | +| `nexent_chain_of_diagnosis` | **自定义** | CoD诊断推理链 | +| `nexent_evaluate_diagnosis_confidence` | **自定义** | 置信度评估 | +| `nexent_start_diagnosis_game` | **自定义** | 启动诊断游戏 | +| `nexent_diagnosis_action` | **自定义** | 诊断游戏动作 | +| `nexent_search_pathology_images` | **自定义** | 病理图片搜索 | +| `nexent_generate_medical_guide` | **自定义** | 就医指南生成 | + +## Prompt 配置 + +### duty_prompt (角色提示词) + +``` +# 🏥 病理学AI助手 + +你是一位专业的病理学AI助手。 + +## ⚠️ 最重要规则 + +### 1. 双重检索（必须执行） +回答医学问题前，必须同时调用： +- knowledge_base_search(query="关键词", search_mode="hybrid") +- tavily_search(query="关键词") + +权重：内部60% + 外部40% + +### 2. 按钮格式规则 +工具返回的 [btn:xxx] 格式必须原样保留！ + +## 🎮 诊断模拟游戏规则 +1. 每执行一步后必须停止，等待用户选择 +2. 原样输出工具返回的按钮 +3. 不要自己做决定 + +## 安全提醒 +⚠️ 本AI仅供参考，不能替代专业医生诊断。 +``` + +## 知识库配置 + +| 配置项 | 值 | +|--------|-----| +| 知识库名称 | pathology_knowledge | +| 搜索模式 | hybrid | +| 向量数据库 | Elasticsearch | + +## 外部搜索配置 + +| 配置项 | 值 | +|--------|-----| +| 搜索引擎 | Tavily | +| 权重 | 40% | diff --git a/pathology-ai/architecture.md b/pathology-ai/architecture.md new file mode 100644 index 000000000..64a215c68 --- /dev/null +++ b/pathology-ai/architecture.md @@ -0,0 +1,308 @@ +# 🏗️ 架构与调用关系图 + +## 系统架构概览 + +``` +┌─────────────────────────────────────────────────────────────────┐ +│ 用户界面 (Frontend) │ +│ ┌─────────────┐ ┌─────────────┐ ┌─────────────────────────┐ │ +│ │ 聊天界面 │ │ 医学可视化 │ │ 诊断模拟游戏界面 │ │ +│ └─────────────┘ └─────────────┘ └─────────────────────────┘ │ +└────────────────────────────┬────────────────────────────────────┘ + │ + ▼ +┌─────────────────────────────────────────────────────────────────┐ +│ Nexent Runtime │ +│ ┌─────────────────────────────────────────────────────────────┐│ +│ │ 病理学AI助手 (Agent ID: 13) ││ +│ └─────────────────────────────────────────────────────────────┘│ +└────────────────────────────┬────────────────────────────────────┘ + │ + ▼ +┌─────────────────────────────────────────────────────────────────┐ +│ MCP 工具层 │ +│ ┌──────────────────────┐ ┌──────────────────────────────┐ │ +│ │ 内置工具 │ │ 自定义医疗工具 │ │ +│ │ • knowledge_search │ │ • chain_of_diagnosis │ │ +│ │ • tavily_search │ │ • evaluate_confidence │ │ +│ │ • analyze_image │ │ • diagnosis_game │ │ +│ └──────────────────────┘ └──────────────────────────────┘ │ +└────────────────────────────┬────────────────────────────────────┘ + │ + ▼ +┌─────────────────────────────────────────────────────────────────┐ +│ 数据层 │ +│ ┌─────────────┐ ┌─────────────┐ ┌─────────────────────────┐ │ +│ │ PostgreSQL │ │Elasticsearch│ │ 病理图片服务器 │ │ +│ └─────────────┘ └─────────────┘ └─────────────────────────┘ │ +└─────────────────────────────────────────────────────────────────┘ +``` + +--- + +## 工具调用关系图 (Mermaid) + +```mermaid +flowchart TD + subgraph User["👤 用户"] + Q[用户提问] + end + + subgraph Agent["🤖 病理学AI助手 Agent ID:13"] + A[接收问题] + B{判断问题类型} + end + + subgraph BuiltIn["📦 内置工具"] + KB[knowledge_base_search
内部知识库 60%] + TS[tavily_search
外部搜索 40%] + AI[analyze_image
图片分析] + end + + subgraph Medical["🏥 医疗诊断工具"] + COD[chain_of_diagnosis
5步诊断推理链] + CONF[evaluate_diagnosis_confidence
置信度评估] + IMG[search_pathology_images
病理图片搜索] + GUIDE[generate_medical_guide
就医指南] + end + + subgraph Game["🎮 诊断模拟游戏"] + START[start_diagnosis_game
启动游戏] + ACTION[diagnosis_action
问诊/体检/检查/诊断] + end + + subgraph Visual["📊 医学可视化工具"] + KG[generate_knowledge_graph
知识图谱] + FLOW[generate_diagnosis_flow
诊断流程图] + CHART[generate_medical_chart
统计图表] + RADAR[generate_radar_chart
雷达图] + TL[generate_timeline
时间线] + GANTT[generate_gantt_chart
甘特图] + QUAD[generate_quadrant_chart
象限图] + STATE[generate_state_diagram
状态图] + SANKEY[generate_sankey_diagram
桑基图] + end + + Q --> A + A --> B + B -->|医学问答| KB + B -->|医学问答| TS + B -->|图片分析| AI + B -->|诊断分析| COD + B -->|诊断游戏| START + B -->|可视化| KG + B -->|就医咨询| GUIDE + + KB --> R[生成回答] + TS --> R + AI --> R + COD --> CONF --> R + START --> ACTION --> R + KG --> R + FLOW --> R + CHART --> R + RADAR --> R + TL --> R + GANTT --> R + QUAD --> R + STATE --> R + SANKEY --> R + IMG --> R + GUIDE --> R +``` + +--- + +## 完整工具调用流程图 + +```mermaid +flowchart LR + subgraph Input["输入"] + U[用户问题] + end + + subgraph Process["处理流程"] + U --> Agent[病理学AI助手] + Agent --> Parse{解析意图} + + Parse -->|知识查询| Search["双重检索"] + Parse -->|诊断请求| Diag["诊断分析"] + Parse -->|游戏请求| Game["诊断游戏"] + Parse -->|可视化| Viz["图表生成"] + Parse -->|就医咨询| Guide["就医指南"] + + Search --> KB[内部知识库] + Search --> Web[外部搜索] + + Diag --> CoD[CoD推理链] + CoD --> Eval[置信度评估] + + Game --> Start[启动游戏] + Start --> Action[执行动作] + + Viz --> Charts[9种图表工具] + end + + subgraph Output["输出"] + KB --> Merge[结果融合] + Web --> Merge + Eval --> Merge + Action --> Merge + Charts --> Merge + Guide --> Merge + Merge --> Response[AI回答] + end +``` + +--- + +## 诊断游戏流程图 + +```mermaid +flowchart LR + A[启动游戏] --> B[问诊阶段] + B --> C[体格检查] + C --> D[辅助检查] + D --> E[给出诊断] + E --> F[评分反馈] +``` + +--- + +## Chain-of-Diagnosis 流程 + +```mermaid +flowchart TD + INPUT[输入症状] --> S1[Step1: 症状分析] + S1 --> S2[Step2: 病史关联] + S2 --> S3[Step3: 鉴别诊断] + S3 --> S4[Step4: 检查建议] + S4 --> S5[Step5: 初步结论] + S5 --> EVAL[置信度评估] + EVAL --> OUTPUT[诊断报告] +``` + +--- + +## 前端组件调用关系 + +```mermaid +flowchart TD + subgraph Chat["聊天界面"] + CI[chatInterface.tsx] + MD[markdownRenderer.tsx] + end + + subgraph MedViz["医学可视化组件"] + MVP[MedicalVisualizationPanel] + PIG[PathologyImageGallery] + DCC[DiagnosisConfidenceCard] + ST[SourceTag] + end + + subgraph Services["服务层"] + CS[conversationService.ts] + end + + CI --> MD + MD -->|渲染Mermaid| MVP + MD -->|渲染图片| PIG + MD -->|渲染置信度| DCC + MD -->|渲染来源| ST + MD -->|[btn:xx]按钮| BTN[ClickableOption] + + CI --> CS + CS -->|deleteAll| API[后端API] +``` + +--- + +## MCP工具注册关系 + +```mermaid +flowchart TD + subgraph MCP["FastMCP框架"] + LMS[local_mcp_service.py] + end + + subgraph Decorator["@local_mcp_service.tool 装饰器"] + D1[医疗诊断工具 x4] + D2[诊断游戏工具 x2] + D3[可视化工具 x9] + end + + subgraph Runtime["Nexent Runtime"] + Agent[病理学AI助手] + end + + LMS --> D1 + LMS --> D2 + LMS --> D3 + + D1 --> Agent + D2 --> Agent + D3 --> Agent +``` + +--- + +## 数据流向图 + +```mermaid +flowchart LR + subgraph Input["用户输入"] + Text[文本问题] + Image[病理图片] + Click[按钮点击] + end + + subgraph Process["Agent处理"] + Text --> Agent[病理学AI助手] + Image --> Agent + Click --> Agent + + Agent --> Tools[MCP工具] + Tools --> KB[(知识库)] + Tools --> Web((互联网)) + Tools --> ImgDB[(图片库)] + end + + subgraph Output["输出"] + KB --> Response + Web --> Response + ImgDB --> Response + Response[AI回答] --> Render[前端渲染] + Render --> Markdown[文本/表格] + Render --> Mermaid[图表] + Render --> Images[图片] + Render --> Buttons[交互按钮] + end +``` + +--- + +## 文件修改清单 + +### 后端修改 + +| 文件 | 类型 | 说明 | +|------|------|------| +| `backend/tool_collection/mcp/local_mcp_service.py` | 新增 | 15个医疗MCP工具 | + +### 前端修改 + +| 文件 | 类型 | 说明 | +|------|------|------| +| `frontend/components/medical-visualization/PathologyImageGallery.tsx` | 新增 | 病理图片画廊 | +| `frontend/components/medical-visualization/DiagnosisConfidenceCard.tsx` | 新增 | 置信度卡片 | +| `frontend/components/medical-visualization/SourceTag.tsx` | 新增 | 来源标签 | +| `frontend/components/medical-visualization/MedicalVisualizationPanel.tsx` | 修改 | 去除硬编码 | +| `frontend/components/ui/markdownRenderer.tsx` | 修改 | [btn:xx]按钮解析 | +| `frontend/app/[locale]/chat/components/chatLeftSidebar.tsx` | 修改 | 清空对话按钮 | +| `frontend/services/conversationService.ts` | 修改 | deleteAll方法 | + +### 配置文件 + +| 文件 | 说明 | +|------|------| +| `docker/update_prompt_btn.sql` | Agent提示词配置 | diff --git a/pathology-ai/code-changes/backend/local_mcp_service.py b/pathology-ai/code-changes/backend/local_mcp_service.py new file mode 100644 index 000000000..de4421f99 --- /dev/null +++ b/pathology-ai/code-changes/backend/local_mcp_service.py @@ -0,0 +1,1425 @@ +from fastmcp import FastMCP +import json +import re +from typing import Optional, List, Dict +from dataclasses import dataclass, field +from enum import Enum + +# Create MCP server +local_mcp_service = FastMCP("nexent") + +# ============ Medical Extension Classes ============ + +class ConfidenceLevel(Enum): + """置信度等级""" + HIGH = "HIGH" # >85% 高置信度 + MEDIUM = "MEDIUM" # 60-85% 中等置信度 + LOW = "LOW" # <60% 低置信度 + UNCERTAIN = "UNCERTAIN" # 不确定 + +class RiskLevel(Enum): + """风险等级""" + CRITICAL = "critical" + HIGH = "high" + MEDIUM = "medium" + LOW = "low" + +@local_mcp_service.tool(name="test_tool_name", + description="test_tool_description") +async def demo_tool(para_1: str, para_2: int) -> str: + print("tool is called successfully") + print(para_1, para_2) + return "success" + + +# ============ Medical Visualization Tools (Dynamic Generation) ============ + +@local_mcp_service.tool( + name="generate_knowledge_graph", + description="""生成医学知识图谱(Mermaid flowchart格式)。 + +参数说明: +- topic: 图谱主题 +- nodes: 节点列表，用|分隔，格式为"节点1|节点2|节点3" +- relations: 关系列表，用|分隔，格式为"节点1-->节点2|节点2-->节点3" + +使用方法: 先用知识库搜索获取相关概念，然后提取关键概念作为nodes，概念间的关系作为relations传入此工具。""" +) +async def generate_knowledge_graph(topic: str, nodes: str = "", relations: str = "") -> str: + """Generate dynamic knowledge graph based on provided nodes and relations""" + + # Parse nodes and relations + node_list = [n.strip() for n in nodes.split("|") if n.strip()] if nodes else [] + relation_list = [r.strip() for r in relations.split("|") if r.strip()] if relations else [] + + # If no nodes provided, return instruction + if not node_list: + return f"""请先使用知识库搜索获取关于"{topic}"的相关信息，然后提取关键概念和关系，再调用此工具。 + +示例调用: +generate_knowledge_graph( + topic="HIV感染机制", + nodes="HIV病毒|CD4细胞|免疫系统|病毒复制|机会性感染", + relations="HIV病毒-->CD4细胞|CD4细胞-->免疫系统|HIV病毒-->病毒复制|免疫系统-->机会性感染" +)""" + + # Create node map + node_map = {node: f"N{i}" for i, node in enumerate(node_list)} + + # Parse relations and find root nodes (nodes that are sources but not targets) + sources = set() + targets = set() + parsed_relations = [] + for rel in relation_list: + if "-->" in rel: + parts = rel.split("-->") + if len(parts) == 2: + src, tgt = parts[0].strip(), parts[1].strip() + if src in node_map and tgt in node_map: + sources.add(src) + targets.add(tgt) + parsed_relations.append((src, tgt)) + + # Group nodes by level (simple BFS-like grouping) + root_nodes = [n for n in node_list if n in sources and n not in targets] + if not root_nodes: + root_nodes = [node_list[0]] if node_list else [] + + # Build hierarchical layout using subgraphs + lines = ["flowchart TB"] + + # Add all nodes with rounded rectangle style + for i, node in enumerate(node_list): + node_id = node_map[node] + lines.append(f' {node_id}(["{node}"])') + + # Add relations with labels + for src, tgt in parsed_relations: + lines.append(f" {node_map[src]} --> {node_map[tgt]}") + + # Add gradient colors for better visual + gradient_colors = [ + "#667eea", "#764ba2", "#f093fb", "#f5576c", + "#4facfe", "#00f2fe", "#43e97b", "#38f9d7", + "#fa709a", "#fee140", "#a8edea", "#fed6e3" + ] + for i, node in enumerate(node_list): + color = gradient_colors[i % len(gradient_colors)] + lines.append(f" style {node_map[node]} fill:{color},color:#fff,stroke:{color},stroke-width:2px") + + # Add link styles + lines.append(" linkStyle default stroke:#666,stroke-width:2px") + + mermaid_code = f'''```mermaid +{chr(10).join(lines)} +```''' + + return mermaid_code + + +@local_mcp_service.tool( + name="generate_diagnosis_flow", + description="""生成诊断流程图(Mermaid flowchart格式)。 + +参数说明: +- disease: 疾病名称 +- steps: 流程步骤列表，用|分隔，格式为"步骤1|步骤2|步骤3" +- decisions: 决策点列表，用|分隔，格式为"决策1:是选项,否选项|决策2:选项A,选项B" + +使用方法: 根据知识库搜索结果，提取诊断流程的关键步骤和决策点。""" +) +async def generate_diagnosis_flow(disease: str, steps: str = "", decisions: str = "") -> str: + """Generate compact horizontal diagnosis flowchart""" + + step_list = [s.strip() for s in steps.split("|") if s.strip()] if steps else [] + + if not step_list: + return f"""请搜索"{disease}"诊断流程，提取关键步骤。 +示例: steps="初筛|确证|检测|治疗" """ + + # Horizontal layout (left to right) - reduces height + lines = ["flowchart LR"] + + # Full node names, horizontal flow + for i, step in enumerate(step_list): + node_id = f"S{i}" + + if i == 0: + lines.append(f' {node_id}(("{step}"))') + elif i == len(step_list) - 1: + lines.append(f' {node_id}(("{step}"))') + else: + lines.append(f' {node_id}["{step}"]') + + # Connect all nodes + node_chain = " --> ".join([f"S{i}" for i in range(len(step_list))]) + lines.append(f" {node_chain}") + + # Gradient colors + colors = ["#6366f1", "#8b5cf6", "#a855f7", "#ec4899", "#f43f5e", "#f97316", "#22c55e"] + for i in range(len(step_list)): + color = colors[i % len(colors)] + lines.append(f" style S{i} fill:{color},color:#fff,stroke:#fff") + + mermaid_code = f'''```mermaid +{chr(10).join(lines)} +```''' + + return mermaid_code + + +@local_mcp_service.tool( + name="generate_medical_chart", + description="""生成统计图表(Mermaid格式)。 + +参数说明: +- chart_type: 图表类型 - pie(饼图), bar(柱状图), line(折线图) +- title: 图表标题 +- labels: 标签列表，用|分隔，如"类别1|类别2|类别3" +- values: 数值列表，用|分隔，如"30|25|45" + +使用方法: 根据数据分析结果，提取分类和数值传入此工具。""" +) +async def generate_medical_chart(chart_type: str, title: str, labels: str = "", values: str = "") -> str: + """Generate dynamic statistics chart""" + + label_list = [l.strip() for l in labels.split("|") if l.strip()] if labels else [] + value_list = [v.strip() for v in values.split("|") if v.strip()] if values else [] + + if not label_list or not value_list: + return f"""请提供数据的标签和数值。 + +示例调用: +generate_medical_chart( + chart_type="pie", + title="知识分类分布", + labels="病理机制|临床表现|诊断检测|治疗方案", + values="35|25|20|20" +)""" + + if chart_type == "pie": + pie_data = "\n".join([f' "{label}" : {value}' for label, value in zip(label_list, value_list)]) + mermaid_code = f'''```mermaid +pie showData title {title} +{pie_data} +```''' + elif chart_type == "bar": + mermaid_code = f'''```mermaid +xychart-beta + title "{title}" + x-axis [{", ".join(label_list)}] + y-axis "数量" 0 --> {int(max([int(v) for v in value_list]) * 1.2)} + bar [{", ".join(value_list)}] +```''' + elif chart_type == "line": + mermaid_code = f'''```mermaid +xychart-beta + title "{title}" + x-axis [{", ".join(label_list)}] + y-axis "数值" 0 --> {int(max([int(v) for v in value_list]) * 1.2)} + line [{", ".join(value_list)}] +```''' + else: + mermaid_code = f"不支持的图表类型: {chart_type}。请使用 pie, bar, 或 line" + + return mermaid_code + + +# ============ Advanced Medical Visualization Tools ============ + +@local_mcp_service.tool( + name="generate_radar_chart", + description="""生成雷达图/蛛网图，用于多维度健康指标对比分析。 + +参数说明: +- title: 图表标题 +- dimensions: 维度列表，用|分隔，如"指标1|指标2|指标3|指标4|指标5" +- values: 数值列表(0-100)，用|分隔，如"80|65|90|75|85" +- benchmark: 可选，基准值列表，用于对比 + +应用场景: 健康评估、症状严重程度评分、治疗效果多维对比""" +) +async def generate_radar_chart(title: str, dimensions: str = "", values: str = "", benchmark: str = "") -> str: + """Generate radar/spider chart for multi-dimensional comparison""" + + dim_list = [d.strip() for d in dimensions.split("|") if d.strip()] if dimensions else [] + val_list = [v.strip() for v in values.split("|") if v.strip()] if values else [] + + if not dim_list or not val_list or len(dim_list) < 3: + return f"""雷达图需要至少3个维度。 + +示例调用: +generate_radar_chart( + title="HIV患者健康评估", + dimensions="免疫功能|病毒载量|肝功能|肾功能|心血管|神经系统", + values="75|60|85|90|80|70" +)""" + + # 使用quadrantChart模拟雷达图效果，或用表格+描述替代 + # Mermaid暂不直接支持雷达图，用可视化描述+数据表格代替 + + # 生成数据可视化表格 + table_rows = [] + for dim, val in zip(dim_list, val_list): + val_int = int(val) if val.isdigit() else 50 + bar = "█" * (val_int // 10) + "░" * (10 - val_int // 10) + status = "🟢" if val_int >= 80 else "🟡" if val_int >= 60 else "🔴" + table_rows.append(f"| {dim} | {bar} | {val}% | {status} |") + + result = f"""### 📊 {title} + +| 评估维度 | 指标条形图 | 数值 | 状态 | +|---------|-----------|------|------| +{chr(10).join(table_rows)} + +**评估说明:** 🟢优秀(≥80) 🟡良好(60-79) 🔴需关注(<60) + +```mermaid +pie showData title {title} +{chr(10).join([f' "{dim}" : {val}' for dim, val in zip(dim_list, val_list)])} +```""" + + return result + + +@local_mcp_service.tool( + name="generate_medical_guide", + description="""生成清晰的就医指南，包含就医方式选择和就医流程。 + +参数说明: +- condition: 病情描述(如"HIV患者呼吸困难") +- urgency: 紧急程度(emergency/urgent/routine) +- patient_info: 患者关键信息(如"CD4计数150") + +返回格式化的就医指南，包含多种就医方式和详细流程。""" +) +async def generate_medical_guide(condition: str, urgency: str = "urgent", patient_info: str = "") -> str: + """Generate formatted medical guide""" + + urgency_map = { + "emergency": ("🚨 紧急", "立即拨打120"), + "urgent": ("⚠️ 紧急", "尽快就医"), + "routine": ("📋 常规", "预约就诊"), + } + + urgency_label, urgency_action = urgency_map.get(urgency, ("⚠️ 紧急", "尽快就医")) + + guide = f"""# 🏥 就医指南 + +## 📋 病情概述 +- **症状**: {condition} +- **患者信息**: {patient_info if patient_info else "未提供"} +- **紧急程度**: {urgency_label} + +--- + +## 🚗 就医方式选择 + +### 方式1: 拨打120 {"✅ 推荐" if urgency == "emergency" else ""} + +| 步骤 | 操作 | +|------|------| +| 1️⃣ | 拨打120急救电话 | +| 2️⃣ | 告知: {condition}，{patient_info if patient_info else "病情紧急"} | +| 3️⃣ | 告知当前位置，等待救护车 | +| 4️⃣ | 由医护人员送往医院 | + +### 方式2: 自行前往医院 {"✅ 推荐" if urgency == "urgent" else ""} + +| 步骤 | 操作 | +|------|------| +| 1️⃣ | 选择最近的三甲医院 | +| 2️⃣ | 电话或微信预约挂号(急诊) | +| 3️⃣ | 由家属陪同前往 | +| 4️⃣ | 直接进入急诊科 | + +### 方式3: 拨打医院急诊科 + +| 步骤 | 操作 | +|------|------| +| 1️⃣ | 拨打目标医院总机 | +| 2️⃣ | 转接急诊科说明病情 | +| 3️⃣ | 按指导前往医院 | + +--- + +## 🏥 到院后流程 + +```mermaid +flowchart LR + A[到达医院] --> B[挂号/急诊登记] + B --> C[初诊评估] + C --> D[体格检查] + D --> E[辅助检查] + E --> F[诊断确认] + F --> G[治疗方案] + G --> H[住院/出院] +``` + +### 详细步骤 + +| 序号 | 环节 | 具体内容 | +|------|------|----------| +| 1 | **登记** | 挂号/急诊登记，说明{condition} | +| 2 | **初诊** | 医生问诊，测量生命体征 | +| 3 | **体检** | 听诊肺部等体格检查 | +| 4 | **检查** | 胸部X光/CT、血液检查、血气分析 | +| 5 | **诊断** | 等待结果(通常24-48小时) | +| 6 | **治疗** | 制定方案，开始治疗 | +| 7 | **监测** | 监测疗效和不良反应 | + +--- + +## ⚠️ 注意事项 + +- 携带身份证、医保卡 +- 携带既往病历和检查报告 +- 如有HIV相关资料请一并携带 +- 保持通讯畅通 + +> 💡 **提示**: {urgency_action}，不要延误治疗时机 +""" + + return guide + + +@local_mcp_service.tool( + name="generate_timeline", + description="""生成时间线图，用于展示疾病发展历程或治疗计划。 + +参数说明: +- title: 时间线标题 +- events: 事件列表，用|分隔，格式为"时间点:事件描述|时间点:事件描述" + +应用场景: 病程发展、治疗时间线、随访计划""" +) +async def generate_timeline(title: str, events: str = "") -> str: + """Generate timeline diagram""" + + event_list = [e.strip() for e in events.split("|") if e.strip()] if events else [] + + if not event_list: + return f"""请提供时间线事件。 + +示例调用: +generate_timeline( + title="HIV感染自然病程", + events="感染期:HIV病毒侵入|急性期:病毒快速复制|潜伏期:免疫平衡|AIDS期:免疫崩溃" +)""" + + lines = ["timeline", f" title {title}"] + + for event in event_list: + if ":" in event: + time_point, description = event.split(":", 1) + lines.append(f" {time_point.strip()}") + lines.append(f" : {description.strip()}") + else: + lines.append(f" {event}") + + mermaid_code = f'''```mermaid +{chr(10).join(lines)} +```''' + + return mermaid_code + + +@local_mcp_service.tool( + name="generate_gantt_chart", + description="""生成甘特图，用于治疗计划和疗程安排。 + +参数说明: +- title: 图表标题 +- tasks: 任务列表，用|分隔，格式为"任务名:开始日期,持续天数|任务名:开始日期,持续天数" + +应用场景: 治疗方案安排、康复计划、随访时间表""" +) +async def generate_gantt_chart(title: str, tasks: str = "") -> str: + """Generate Gantt chart for treatment planning""" + + task_list = [t.strip() for t in tasks.split("|") if t.strip()] if tasks else [] + + if not task_list: + return f"""请提供治疗任务安排。 + +示例调用: +generate_gantt_chart( + title="HIV抗病毒治疗计划", + tasks="初始评估:2024-01-01,7d|药物启动:2024-01-08,30d|首次复查:2024-02-07,1d|稳定期治疗:2024-02-08,90d" +)""" + + lines = [ + "gantt", + f" title {title}", + " dateFormat YYYY-MM-DD", + " section 治疗阶段" + ] + + for i, task in enumerate(task_list): + if ":" in task: + task_name, timing = task.split(":", 1) + if "," in timing: + start_date, duration = timing.split(",", 1) + lines.append(f" {task_name.strip()} : t{i}, {start_date.strip()}, {duration.strip()}") + else: + lines.append(f" {task_name.strip()} : t{i}, {timing.strip()}") + else: + lines.append(f" {task} : t{i}, 7d") + + mermaid_code = f'''```mermaid +{chr(10).join(lines)} +```''' + + return mermaid_code + + +@local_mcp_service.tool( + name="generate_quadrant_chart", + description="""生成象限图，用于风险评估和优先级分析。 + +参数说明: +- title: 图表标题 +- x_axis: X轴标签(低到高) +- y_axis: Y轴标签(低到高) +- items: 项目列表，格式为"项目名:x坐标,y坐标|项目名:x坐标,y坐标" (坐标范围0-1) + +应用场景: 疾病风险评估、治疗优先级、药物选择矩阵""" +) +async def generate_quadrant_chart(title: str, x_axis: str = "紧急程度", y_axis: str = "重要程度", items: str = "") -> str: + """Generate quadrant chart for risk assessment""" + + item_list = [i.strip() for i in items.split("|") if i.strip()] if items else [] + + if not item_list: + return f"""请提供评估项目。 + +示例调用: +generate_quadrant_chart( + title="HIV并发症处理优先级", + x_axis="紧急程度", + y_axis="严重程度", + items="机会性感染:0.9,0.85|肝功能异常:0.6,0.7|皮疹反应:0.4,0.3|轻度贫血:0.2,0.4" +)""" + + lines = [ + "quadrantChart", + f" title {title}", + f' x-axis "低{x_axis}" --> "高{x_axis}"', + f' y-axis "低{y_axis}" --> "高{y_axis}"', + ' quadrant-1 "紧急重要"', + ' quadrant-2 "重要不紧急"', + ' quadrant-3 "不重要不紧急"', + ' quadrant-4 "紧急不重要"' + ] + + for item in item_list: + if ":" in item: + name, coords = item.split(":", 1) + if "," in coords: + x, y = coords.split(",", 1) + lines.append(f' "{name.strip()}": [{x.strip()}, {y.strip()}]') + + mermaid_code = f'''```mermaid +{chr(10).join(lines)} +```''' + + return mermaid_code + + +@local_mcp_service.tool( + name="generate_state_diagram", + description="""生成状态转换图，用于展示疾病状态变化。 + +参数说明: +- title: 图表标题 +- states: 状态列表，用|分隔 +- transitions: 转换列表，格式为"状态1-->状态2:触发条件|状态2-->状态3:触发条件" + +应用场景: 疾病分期、病情演变、治疗响应状态""" +) +async def generate_state_diagram(title: str, states: str = "", transitions: str = "") -> str: + """Generate state diagram for disease progression""" + + state_list = [s.strip() for s in states.split("|") if s.strip()] if states else [] + trans_list = [t.strip() for t in transitions.split("|") if t.strip()] if transitions else [] + + if not state_list or not trans_list: + return f"""请提供状态和转换关系。 + +示例调用: +generate_state_diagram( + title="HIV感染分期", + states="健康|急性感染|临床潜伏期|AIDS期", + transitions="健康-->急性感染:HIV暴露|急性感染-->临床潜伏期:免疫应答|临床潜伏期-->AIDS期:CD4<200" +)""" + + lines = ["stateDiagram-v2"] + + # Add state descriptions + state_map = {s: f"s{i}" for i, s in enumerate(state_list)} + for state, sid in state_map.items(): + lines.append(f' {sid} : {state}') + + # Add transitions + for trans in trans_list: + if "-->" in trans: + parts = trans.split("-->") + if len(parts) == 2: + src = parts[0].strip() + tgt_label = parts[1] + if ":" in tgt_label: + tgt, label = tgt_label.split(":", 1) + tgt = tgt.strip() + if src in state_map and tgt in state_map: + lines.append(f' {state_map[src]} --> {state_map[tgt]} : {label.strip()}') + else: + tgt = tgt_label.strip() + if src in state_map and tgt in state_map: + lines.append(f' {state_map[src]} --> {state_map[tgt]}') + + # Mark start and end + if state_list: + lines.insert(1, f' [*] --> {state_map[state_list[0]]}') + lines.append(f' {state_map[state_list[-1]]} --> [*]') + + mermaid_code = f'''```mermaid +{chr(10).join(lines)} +```''' + + return mermaid_code + + +@local_mcp_service.tool( + name="generate_sankey_diagram", + description="""生成桑基图，用于展示流量和转换关系。 + +参数说明: +- title: 图表标题 +- flows: 流向列表，格式为"源,目标,数量|源,目标,数量" + +应用场景: 诊断分流、患者转归、治疗路径""" +) +async def generate_sankey_diagram(title: str, flows: str = "") -> str: + """Generate Sankey diagram for flow visualization""" + + flow_list = [f.strip() for f in flows.split("|") if f.strip()] if flows else [] + + if not flow_list: + return f"""请提供流向数据。 + +示例调用: +generate_sankey_diagram( + title="HIV筛查诊断流程", + flows="初筛人群,阳性,150|初筛人群,阴性,850|阳性,确证阳性,140|阳性,假阳性,10|确证阳性,入组治疗,130|确证阳性,暂缓治疗,10" +)""" + + lines = ["sankey-beta", ""] + + for flow in flow_list: + parts = flow.split(",") + if len(parts) >= 3: + src, tgt, val = parts[0].strip(), parts[1].strip(), parts[2].strip() + lines.append(f'{src},{tgt},{val}') + + mermaid_code = f'''```mermaid +{chr(10).join(lines)} +``` + +**{title}** - 流向分析图''' + + return mermaid_code + + +# ============ 诊断模拟器 - 医学教育游戏化 ============ + +import random + +# 预设病例库 +CASE_LIBRARY = { + "hiv_basic": { + "patient": "李先生，32岁，已婚", + "chief_complaint": "反复发热、乏力1个月", + "history": { + "发热情况": "低热为主，体温37.5-38.2℃，午后明显", + "其他症状": "明显乏力，体重下降约5kg", + "既往史": "既往体健，无慢性病史", + "接触史": "3个月前有不洁性行为史", + "用药情况": "自行服用退烧药，效果不佳" + }, + "physical_exam": { + "一般情况": "神志清楚，精神欠佳，消瘦", + "淋巴结": "颈部、腋窝淋巴结肿大，无压痛", + "口腔": "可见口腔白斑", + "皮肤": "无皮疹" + }, + "lab_tests": { + "血常规": "WBC 3.2×10^9/L↓，淋巴细胞比例降低", + "HIV抗体初筛": "阳性", + "HIV确证试验": "阳性", + "CD4计数": "186个/μL↓↓", + "病毒载量": "125,000 copies/mL" + }, + "diagnosis": "HIV感染/AIDS期", + "difficulty": 1, + "key_points": ["接触史询问", "淋巴结检查", "HIV筛查", "CD4计数"] + }, + "hiv_opportunistic": { + "patient": "王女士，45岁", + "chief_complaint": "咳嗽、气促2周，加重3天", + "history": { + "呼吸症状": "干咳为主，活动后气促明显", + "发热情况": "持续低热，夜间盗汗", + "既往史": "HIV感染史5年，未规律服药", + "用药情况": "间断服用抗病毒药物" + }, + "physical_exam": { + "一般情况": "呼吸急促，口唇轻度发绀", + "肺部": "双肺呼吸音粗，可闻及少量湿啰音", + "口腔": "舌面白色斑块，可刮除" + }, + "lab_tests": { + "血气分析": "PaO2 58mmHg↓", + "CD4计数": "45个/μL↓↓↓", + "胸部CT": "双肺弥漫性磨玻璃影", + "痰检": "六胺银染色见肺孢子菌" + }, + "diagnosis": "AIDS合并肺孢子菌肺炎(PCP)", + "difficulty": 2, + "key_points": ["服药依从性", "机会性感染识别", "CD4与感染风险"] + } +} + +@local_mcp_service.tool( + name="start_diagnosis_game", + description="""启动诊断模拟游戏。用户扮演医生，AI扮演患者，进行问诊练习。 + +参数说明: +- difficulty: 难度等级 (1=初级, 2=中级, 3=高级) +- case_type: 病例类型，可选 "hiv_basic"(HIV基础), "hiv_opportunistic"(机会性感染), "random"(随机) + +游戏流程: 问诊→体检→检查→诊断，最终给出评分""" +) +async def start_diagnosis_game(difficulty: int = 1, case_type: str = "random") -> str: + """Start an interactive diagnosis simulation game""" + + # 选择病例 + if case_type == "random" or case_type not in CASE_LIBRARY: + case_key = random.choice(list(CASE_LIBRARY.keys())) + else: + case_key = case_type + + case = CASE_LIBRARY[case_key] + + result = f""" +## 🏥 诊断模拟器 - 病例开始 + +### 👤 患者信息 +**{case['patient']}** + +### 💬 主诉 +> "{case['chief_complaint']}" + +--- + +### 📋 当前阶段：问诊 (第1步/共4步) + +**请选择您要询问的内容：** + +[btn:询问发热详情] [btn:询问其他症状] [btn:询问既往病史] +[btn:询问接触史] [btn:询问用药情况] [btn:进入体格检查] + +💡 **提示**：全面的问诊是正确诊断的基础，请尽量收集完整病史信息。 + +--- +*难度：{"⭐" * case['difficulty']} | 病例ID：{case_key}* +""" + + return result + + +@local_mcp_service.tool( + name="diagnosis_action", + description="""在诊断模拟中执行动作（问诊/检查/诊断）。 + +参数说明: +- case_id: 病例ID +- action_type: 动作类型 (ask=问诊, exam=体检, test=检查, diagnose=诊断) +- action_detail: 具体动作内容 + +示例: diagnosis_action(case_id="hiv_basic", action_type="ask", action_detail="发热情况")""" +) +async def diagnosis_action(case_id: str, action_type: str, action_detail: str) -> str: + """Process a diagnosis action in the simulation""" + + if case_id not in CASE_LIBRARY: + return "❌ 病例不存在，请先使用 start_diagnosis_game 开始新游戏" + + case = CASE_LIBRARY[case_id] + + if action_type == "ask": + # 问诊阶段 + if action_detail in case["history"]: + response = case["history"][action_detail] + return f""" +### 👤 患者回答 + +**关于{action_detail}：** +> "{response}" + +--- + +**继续问诊或进入下一阶段：** + +[btn:询问发热情况] [btn:询问其他症状] [btn:询问既往史] +[btn:询问接触史] [btn:询问用药情况] [btn:进入体格检查] +""" + else: + return f""" +### 👤 患者回答 + +> "医生，这个...我不太清楚怎么回答。您能换个方式问吗？" + +**可询问的内容：** {', '.join(case['history'].keys())} + +[btn:询问发热情况] [btn:询问其他症状] [btn:询问既往史] +[btn:询问接触史] [btn:询问用药情况] [btn:进入体格检查] +""" + + elif action_type == "exam": + # 体格检查阶段 + if action_detail in case["physical_exam"]: + finding = case["physical_exam"][action_detail] + return f""" +### 🩺 体格检查结果 + +**{action_detail}检查：** +> {finding} + +--- + +**继续检查或进入下一阶段：** + +[btn:检查一般情况] [btn:检查淋巴结] [btn:检查口腔] [btn:检查皮肤] +[btn:开具辅助检查] +""" + else: + return f""" +### 🩺 体格检查 + +该部位检查未见明显异常。 + +**可检查的项目：** {', '.join(case['physical_exam'].keys())} + +[btn:检查一般情况] [btn:检查淋巴结] [btn:检查口腔] [btn:检查皮肤] +[btn:开具辅助检查] +""" + + elif action_type == "test": + # 辅助检查阶段 + if action_detail in case["lab_tests"]: + result = case["lab_tests"][action_detail] + return f""" +### 🔬 检查结果 + +**{action_detail}：** +> {result} + +--- + +**继续检查或给出诊断：** + +[btn:血常规] [btn:HIV抗体初筛] [btn:HIV确证试验] [btn:CD4计数] [btn:病毒载量] +[btn:给出诊断结论] +""" + else: + return f""" +### 🔬 辅助检查 + +该项目暂无结果。 + +**可开具的检查：** {', '.join(case['lab_tests'].keys())} + +[btn:血常规] [btn:HIV抗体初筛] [btn:CD4计数] [btn:病毒载量] +[btn:给出诊断结论] +""" + + elif action_type == "diagnose": + # 诊断阶段 - 评分 + correct = case["diagnosis"].lower() in action_detail.lower() or "hiv" in action_detail.lower() + + if correct: + score = 85 + feedback = "🎉 诊断正确！" + else: + score = 60 + feedback = f"诊断有偏差。正确诊断应为：**{case['diagnosis']}**" + + return f""" +## 🏆 诊断模拟完成！ + +### 您的诊断 +> {action_detail} + +### 标准答案 +> **{case['diagnosis']}** + +--- + +### 📊 评分结果 + +| 评估项目 | 得分 | 说明 | +|---------|------|------| +| 问诊完整度 | 20/25 | 基本覆盖主要病史 | +| 体检针对性 | 22/25 | 检查项目较合理 | +| 辅助检查 | 23/30 | 检查选择恰当 | +| 诊断准确性 | {20 if correct else 10}/20 | {feedback} | + +**总分：{score}/100** {"⭐⭐⭐ 优秀！" if score >= 80 else "⭐⭐ 良好" if score >= 60 else "⭐ 需加强"} + +--- + +### 📚 知识要点回顾 +- **关键线索**：{', '.join(case['key_points'])} +- **诊断依据**：HIV确证试验阳性 + CD4<200 = AIDS期 + +[btn:开始新病例] [btn:查看HIV知识图谱] [btn:返回主页] +""" + + return "未知动作类型，请使用 ask/exam/test/diagnose" + + +# ============ Pathology Image Search Tool ============ + +# 病理图片分类映射 +PATHOLOGY_CATEGORIES = { + "HIV": ["Immunopathology", "Infection"], + "AIDS": ["Immunopathology", "Infection"], + "免疫": ["Immunopathology"], + "感染": ["Infection"], + "心血管": ["Cardiovascular_Pathology", "Atherosclerosis"], + "动脉粥样硬化": ["Atherosclerosis"], + "肺": ["Pulmonary_Pathology"], + "呼吸": ["Pulmonary_Pathology"], + "肿瘤": ["Neoplasia"], + "癌": ["Neoplasia"], + "神经": ["CNS_Pathology"], + "脑": ["CNS_Pathology"], + "胃肠": ["Gastrointestinal_Pathology"], + "消化": ["Gastrointestinal_Pathology"], + "血液": ["Hematopathology"], + "内分泌": ["Endocrine_Pathology"], + "炎症": ["Inflammation"], + "细胞损伤": ["Cell_Injury"], + "电镜": ["Electron_Microscopy"], + "组织学": ["Histology"], +} + +# 每个分类的示例图片及描述 +CATEGORY_IMAGES = { + "Immunopathology": [ + ("0000eb2357e8.jpg", "淋巴细胞浸润，显示免疫反应"), + ("02a4161191a8.jpg", "免疫复合物沉积"), + ("05640ed631c2.jpg", "T细胞介导的免疫损伤"), + ("0f22d896b594.jpg", "B细胞增殖区域"), + ("11a4c1f09706.jpg", "巨噬细胞吞噬活动"), + ], + "Infection": [ + ("075f763add8c.jpg", "病原体感染灶"), + ("0c81a1988e19.jpg", "炎症细胞浸润"), + ("1295dab30912.jpg", "感染性肉芽肿"), + ("17d26c8e5c88.jpg", "组织坏死区域"), + ("22a479f58f04.jpg", "微生物聚集"), + ], + "Cardiovascular_Pathology": [ + ("0606593bb423.jpg", "心肌纤维化"), + ("070dc3e73d66.jpg", "血管内膜增厚"), + ("075032476806.jpg", "心脏瓣膜病变"), + ], + "Atherosclerosis": [ + ("0ba1b0082d67.jpg", "动脉粥样斑块形成"), + ("10474b1d8799.jpg", "脂质沉积"), + ("1575b0d16a3b.jpg", "血管内膜损伤"), + ], + "Pulmonary_Pathology": [ + ("f9f1242c5380.jpg", "肺泡结构改变"), + ("f89a55b691ae.jpg", "支气管炎症"), + ("f7cf9f1ed751.jpg", "肺间质纤维化"), + ], + "Neoplasia": [ + ("00106d3af3f9.jpg", "肿瘤细胞异型性"), + ("0074eed7dc88.jpg", "恶性增殖"), + ("00f1f7a78ea3.jpg", "肿瘤浸润边界"), + ], + "CNS_Pathology": [ + ("021b3f20db2f.jpg", "神经元变性"), + ("02bf3c50f823.jpg", "胶质细胞增生"), + ("083d23ccdd4d.jpg", "脑组织水肿"), + ], + "Gastrointestinal_Pathology": [ + ("00d6f994fc87.jpg", "肠黏膜炎症"), + ("0288d47f9f5b.jpg", "胃溃疡病变"), + ("02a0e46f7c3d.jpg", "肠绒毛萎缩"), + ], + "Hematopathology": [ + ("016b9b2e2cd4.jpg", "骨髓增生"), + ("01e00df21ac8.jpg", "淋巴瘤细胞"), + ("043ce9118f01.jpg", "白血病浸润"), + ], + "Endocrine_Pathology": [ + ("0b21f350e3e9.jpg", "甲状腺滤泡"), + ("0ddee8a2b4f9.jpg", "肾上腺皮质增生"), + ("13cfc5ac2e3b.jpg", "垂体腺瘤"), + ], + "Inflammation": [ + ("00e82b2ec4d0.jpg", "急性炎症反应"), + ("04ad03b22a75.jpg", "慢性炎症浸润"), + ("05eef6d51eaa.jpg", "肉芽组织形成"), + ], + "Cell_Injury": [ + ("063a113740cc.jpg", "细胞水肿"), + ("08672f745e11.jpg", "细胞凋亡"), + ("0d0db3ff6e2f.jpg", "坏死组织"), + ], + "Electron_Microscopy": [ + ("09be997db580.jpg", "细胞超微结构"), + ("0df73df90afe.jpg", "线粒体形态"), + ("1c9d27289d01.jpg", "内质网变化"), + ], + "Histology": [ + ("01b94b8025af.jpg", "正常组织结构"), + ("029bc2eb4a0b.jpg", "细胞形态学"), + ("02c81a6b8380.jpg", "组织切片染色"), + ], +} + +@local_mcp_service.tool( + name="search_pathology_images", + description="""搜索病理学图片。根据关键词返回相关的病理学图片URL。 + +支持的关键词类别: +- HIV/AIDS/免疫: 免疫病理学图片 +- 感染: 感染性疾病图片 +- 心血管/动脉粥样硬化: 心血管病理图片 +- 肺/呼吸: 肺部病理图片 +- 肿瘤/癌: 肿瘤病理图片 +- 神经/脑: 神经系统病理图片 +- 胃肠/消化: 消化系统病理图片 +- 血液: 血液病理图片 +- 炎症: 炎症病理图片 +- 电镜: 电子显微镜图片 +- 组织学: 组织学图片 + +返回Markdown格式的图片，可直接在回复中使用。""" +) +async def search_pathology_images(keyword: str, count: int = 6) -> str: + """Search and return pathology images based on keyword""" + + # 限制返回数量（3的倍数，便于网格布局） + count = min(count, 9) + if count % 3 != 0: + count = (count // 3 + 1) * 3 + + # 查找匹配的分类 + matched_categories = [] + keyword_lower = keyword.lower() + + for key, categories in PATHOLOGY_CATEGORIES.items(): + if key.lower() in keyword_lower or keyword_lower in key.lower(): + matched_categories.extend(categories) + + # 去重 + matched_categories = list(set(matched_categories)) + + if not matched_categories: + # 默认返回免疫病理学图片 + matched_categories = ["Immunopathology", "Infection"] + + # 收集图片信息 (display_url, backend_url, description, category) + image_data = [] + # 前端显示用localhost，后端分析用host.docker.internal + display_base_url = "http://localhost:9012/by_category" + backend_base_url = "http://host.docker.internal:9012/by_category" + + for category in matched_categories: + if category in CATEGORY_IMAGES: + for img_tuple in CATEGORY_IMAGES[category]: + img_file, description = img_tuple + display_url = f"{display_base_url}/{category}/{img_file}" + backend_url = f"{backend_base_url}/{category}/{img_file}" + image_data.append((display_url, backend_url, description, category)) + if len(image_data) >= count: + break + if len(image_data) >= count: + break + + if not image_data: + return f"未找到与'{keyword}'相关的病理图片" + + # 分类名称中文映射 + category_cn = { + "Immunopathology": "免疫病理学", + "Infection": "感染病理学", + "Cardiovascular_Pathology": "心血管病理学", + "Atherosclerosis": "动脉粥样硬化", + "Pulmonary_Pathology": "肺部病理学", + "Neoplasia": "肿瘤病理学", + "CNS_Pathology": "神经病理学", + "Gastrointestinal_Pathology": "消化系统病理学", + "Hematopathology": "血液病理学", + "Endocrine_Pathology": "内分泌病理学", + "Inflammation": "炎症病理学", + "Cell_Injury": "细胞损伤", + "Electron_Microscopy": "电子显微镜", + "Histology": "组织学", + } + + # 生成简洁的Markdown格式 + result = f"## 🔬 {keyword}相关病理图片\n\n" + result += f"已找到 {len(image_data)} 张相关病理学图片：\n\n" + + # 使用简洁的Markdown图片格式 + for i, (display_url, backend_url, desc, cat) in enumerate(image_data, 1): + cat_cn = category_cn.get(cat, cat) + result += f"**{i}. {cat_cn}** - {desc}\n\n" + result += f"\n\n" + + # 提供后端分析用的URL列表（隐藏格式） + backend_urls = [item[1] for item in image_data] + result += f"\n---\n\n" + result += f"📊 **图片来源**: {', '.join([category_cn.get(c, c) for c in matched_categories])}\n\n" + result += f"🔍 **AI分析URL**: `{backend_urls}`\n" + + return result + + +# ============ Chain-of-Diagnosis (CoD) Tool ============ + +# HIV相关知识库 +HIV_KNOWLEDGE = { + "opportunistic_infections": [ + "肺孢子虫肺炎 (PCP)", "巨细胞病毒感染 (CMV)", "隐球菌脑膜炎", + "卡波西肉瘤", "结核病", "弓形虫脑病" + ], + "cd4_thresholds": {"severe": 200, "moderate": 350, "mild": 500}, + "pcp_symptoms": ["干咳", "呼吸困难", "发热", "低氧血症"], + "crypto_symptoms": ["头痛", "发热", "意识改变", "颈强直"], +} + +@local_mcp_service.tool( + name="chain_of_diagnosis", + description="""执行诊断推理链(Chain-of-Diagnosis, CoD)分析。 + +这是一个创新的结构化诊断方法，分5个步骤进行临床推理： +1. 症状分析 - 识别和分析主要症状 +2. 病史关联 - 关联既往病史 +3. 鉴别诊断 - 列出可能的诊断 +4. 检查建议 - 建议进一步检查 +5. 诊断结论 - 给出最终诊断和置信度 + +参数: +- symptoms: 患者症状描述 +- medical_history: 既往病史(可选) +- lab_results: 实验室检查结果(可选) +- imaging_findings: 影像学发现(可选) + +返回结构化的诊断推理报告，包含置信度评估。""" +) +async def chain_of_diagnosis( + symptoms: str, + medical_history: str = "", + lab_results: str = "", + imaging_findings: str = "" +) -> str: + """Execute Chain-of-Diagnosis analysis""" + + reasoning_steps = [] + evidence_collected = [] + + # Step 1: 症状分析 + symptom_analysis = [] + symptom_patterns = { + "呼吸系统": ["咳嗽", "干咳", "呼吸困难", "气短", "胸痛"], + "发热相关": ["发热", "发烧", "高热", "低热"], + "神经系统": ["头痛", "意识改变", "抽搐", "视力改变"], + "消化系统": ["腹泻", "恶心", "呕吐", "腹痛"], + "皮肤表现": ["皮疹", "紫色斑块", "溃疡"], + } + + for system, patterns in symptom_patterns.items(): + found = [p for p in patterns if p in symptoms] + if found: + evidence_collected.extend(found) + symptom_analysis.append(f"{system}: {', '.join(found)}") + + step1_content = "; ".join(symptom_analysis) if symptom_analysis else "症状信息不足" + step1_confidence = 0.8 if evidence_collected else 0.3 + reasoning_steps.append(("症状分析", step1_content, step1_confidence)) + + # Step 2: 病史关联 + history_analysis = "" + is_hiv = False + if medical_history: + if any(kw in medical_history.lower() for kw in ["hiv", "aids", "艾滋", "免疫缺陷"]): + is_hiv = True + history_analysis = "患者有HIV/AIDS病史，需考虑机会性感染" + evidence_collected.append("HIV/AIDS病史") + if any(kw in medical_history for kw in ["免疫抑制", "化疗", "器官移植"]): + history_analysis += "；存在免疫抑制因素" + evidence_collected.append("免疫抑制状态") + + if not history_analysis: + history_analysis = "无特殊病史或病史信息不完整" + + step2_confidence = 0.7 if is_hiv else 0.4 + reasoning_steps.append(("病史关联", history_analysis, step2_confidence)) + + # Step 3: 鉴别诊断 + differentials = [] + cd4_count = None + + if lab_results: + cd4_match = re.search(r'cd4[^\d]*(\d+)', lab_results.lower()) + if cd4_match: + cd4_count = int(cd4_match.group(1)) + evidence_collected.append(f"CD4计数: {cd4_count}") + + if is_hiv: + if cd4_count and cd4_count < 200: + if any(s in symptoms for s in ["干咳", "呼吸困难", "发热"]): + differentials.append("肺孢子虫肺炎 (PCP) - 高度怀疑") + differentials.append("细菌性肺炎") + differentials.append("肺结核") + elif any(s in symptoms for s in ["头痛", "意识"]): + differentials.append("隐球菌脑膜炎") + differentials.append("弓形虫脑病") + else: + differentials.append("需要更多信息进行鉴别") + else: + if any(s in symptoms for s in ["咳嗽", "发热"]): + differentials.extend(["社区获得性肺炎", "病毒性上呼吸道感染", "支气管炎"]) + + step3_content = "鉴别诊断: " + ", ".join(differentials) if differentials else "需要更多信息" + step3_confidence = 0.75 if differentials else 0.3 + reasoning_steps.append(("鉴别诊断", step3_content, step3_confidence)) + + # Step 4: 检查建议 + suggestions = [] + if "PCP" in step3_content or "肺孢子虫" in step3_content: + suggestions = ["诱导痰检查（银染色）", "血气分析", "乳酸脱氢酶(LDH)", "胸部CT"] + elif "脑膜炎" in step3_content: + suggestions = ["腰椎穿刺", "脑脊液墨汁染色", "隐球菌抗原检测", "头颅MRI"] + else: + suggestions = ["血常规", "C反应蛋白", "胸部X线"] + + step4_content = "建议检查: " + ", ".join(suggestions[:4]) + reasoning_steps.append(("检查建议", step4_content, 0.8)) + + # Step 5: 诊断结论 + primary_diagnosis = "诊断待定" + if "高度怀疑" in step3_content: + match = re.search(r'([^,]+)\s*-\s*高度怀疑', step3_content) + if match: + primary_diagnosis = match.group(1).strip() + + step5_content = f"最可能的诊断: {primary_diagnosis}" + step5_confidence = 0.85 if "高度怀疑" in step3_content else 0.5 + reasoning_steps.append(("诊断结论", step5_content, step5_confidence)) + + # 计算总体置信度 + weights = [0.15, 0.15, 0.25, 0.15, 0.30] + overall_confidence = sum(s[2] * w for s, w in zip(reasoning_steps, weights)) + overall_confidence = min(overall_confidence + len(evidence_collected) * 0.02, 1.0) + + # 确定置信度等级 + if overall_confidence >= 0.85: + conf_level = "HIGH" + conf_emoji = "🟢" + elif overall_confidence >= 0.60: + conf_level = "MEDIUM" + conf_emoji = "🟡" + elif overall_confidence >= 0.30: + conf_level = "LOW" + conf_emoji = "🔴" + else: + conf_level = "UNCERTAIN" + conf_emoji = "⚪" + + # 生成报告 + report = "# 🏥 诊断推理链(CoD)分析报告\n\n" + report += "---\n\n" + + for i, (step_name, content, conf) in enumerate(reasoning_steps, 1): + report += f"## 【步骤{i}】{step_name}\n\n" + report += f"{content}\n\n" + report += f"*步骤置信度: {conf*100:.0f}%*\n\n" + + report += "---\n\n" + report += f"## 📊 诊断结果\n\n" + report += f"**主要诊断**: {primary_diagnosis}\n\n" + report += f"**鉴别诊断**: {', '.join([d.split(' - ')[0] for d in differentials if d != primary_diagnosis][:3])}\n\n" + report += f"**置信度**: {conf_emoji} **{conf_level}** ({overall_confidence*100:.1f}%)\n\n" + + # 建议 + report += "## 💡 建议\n\n" + if "PCP" in primary_diagnosis: + report += "- 首选治疗: 复方磺胺甲噁唑 (TMP-SMX)\n" + report += "- 严重病例考虑糖皮质激素辅助治疗\n" + report += "- 监测血氧饱和度\n" + + if conf_level in ["LOW", "UNCERTAIN"]: + report += "- 建议进一步检查以明确诊断\n" + report += "- 必要时请专科会诊\n" + + report += "\n## ⚠️ 重要提示\n\n" + report += "> 本诊断由AI辅助生成，仅供参考。最终诊断请以临床医生判断为准。\n" + + return report + + +# ============ Confidence Evaluation Tool ============ + +@local_mcp_service.tool( + name="evaluate_diagnosis_confidence", + description="""评估诊断的置信度和风险等级。 + +基于证据充分度、一致性、完整性等维度进行量化评估，返回： +- 总体置信度分数和等级 +- 各维度得分 +- 风险等级评估 +- 改进建议 + +参数: +- diagnosis: 诊断结果 +- symptoms: 症状列表，用逗号分隔 +- evidence: 支持证据，用逗号分隔 +- lab_results: 实验室结果(可选)""" +) +async def evaluate_diagnosis_confidence( + diagnosis: str, + symptoms: str = "", + evidence: str = "", + lab_results: str = "" +) -> str: + """Evaluate diagnosis confidence""" + + symptom_list = [s.strip() for s in symptoms.split(",") if s.strip()] + evidence_list = [e.strip() for e in evidence.split(",") if e.strip()] + + # 1. 证据充分度 + evidence_weights = { + "病理确诊": 1.0, "实验室确诊": 0.9, "影像学典型": 0.8, + "临床症状典型": 0.7, "病史支持": 0.6 + } + evidence_score = 0.3 + for e in evidence_list: + for key, weight in evidence_weights.items(): + if key in e: + evidence_score = max(evidence_score, weight) + break + else: + evidence_score += 0.1 + evidence_score = min(evidence_score, 1.0) + + # 2. 一致性评估 + diagnosis_symptom_map = { + "肺孢子虫肺炎": ["干咳", "呼吸困难", "发热"], + "PCP": ["干咳", "呼吸困难", "发热"], + "隐球菌脑膜炎": ["头痛", "发热", "意识改变"], + "肺炎": ["咳嗽", "发热", "胸痛"], + } + consistency_score = 0.5 + for diag_key, expected in diagnosis_symptom_map.items(): + if diag_key in diagnosis: + matched = sum(1 for s in symptom_list if any(es in s for es in expected)) + consistency_score += min(matched / len(expected) * 0.4, 0.4) + break + + # 3. 完整性评估 + completeness_score = 0.0 + if symptom_list: + completeness_score += 0.3 + if evidence_list: + completeness_score += 0.3 + if lab_results: + completeness_score += 0.4 + + # 4. 确定性评估 + certainty_score = 0.5 + uncertain_kw = ["可能", "疑似", "待排除", "考虑"] + certain_kw = ["确诊", "明确", "典型"] + for kw in uncertain_kw: + if kw in diagnosis: + certainty_score -= 0.1 + for kw in certain_kw: + if kw in diagnosis: + certainty_score += 0.15 + certainty_score = max(min(certainty_score, 1.0), 0.1) + + # 计算总体置信度 + weights = {"evidence": 0.35, "consistency": 0.25, "completeness": 0.20, "certainty": 0.20} + overall_score = ( + evidence_score * weights["evidence"] + + consistency_score * weights["consistency"] + + completeness_score * weights["completeness"] + + certainty_score * weights["certainty"] + ) + + # 置信度等级 + if overall_score >= 0.85: + level = "HIGH" + level_emoji = "🟢" + elif overall_score >= 0.60: + level = "MEDIUM" + level_emoji = "🟡" + elif overall_score >= 0.30: + level = "LOW" + level_emoji = "🔴" + else: + level = "UNCERTAIN" + level_emoji = "⚪" + + # 风险等级 + high_risk_kw = ["恶性", "癌", "肿瘤", "急性", "重症", "危重"] + has_high_risk = any(kw in diagnosis for kw in high_risk_kw) + if has_high_risk and overall_score < 0.6: + risk_level = "🔴 CRITICAL" + elif has_high_risk: + risk_level = "🟠 HIGH" + elif overall_score < 0.5: + risk_level = "🟡 MEDIUM" + else: + risk_level = "🟢 LOW" + + # 生成报告 + report = "# 📊 置信度评估报告\n\n" + report += "---\n\n" + report += f"## 总体评估\n\n" + report += f"**诊断**: {diagnosis}\n\n" + report += f"**置信度**: {level_emoji} **{level}** ({overall_score*100:.1f}%)\n\n" + report += f"**风险等级**: {risk_level}\n\n" + + report += "## 📈 各维度得分\n\n" + report += f"| 维度 | 得分 | 说明 |\n" + report += f"|------|------|------|\n" + report += f"| 证据充分度 | {evidence_score*100:.0f}% | 支持诊断的证据质量 |\n" + report += f"| 一致性 | {consistency_score*100:.0f}% | 症状与诊断的匹配度 |\n" + report += f"| 完整性 | {completeness_score*100:.0f}% | 信息的完整程度 |\n" + report += f"| 确定性 | {certainty_score*100:.0f}% | 诊断的明确程度 |\n\n" + + report += "## 💡 改进建议\n\n" + if evidence_score < 0.5: + report += "- 建议补充更多诊断依据\n" + if completeness_score < 0.5: + report += "- 建议完善病史和检查资料\n" + if level in ["LOW", "UNCERTAIN"]: + report += "- 建议进一步检查以明确诊断\n" + report += "- 必要时请专科会诊\n" + if level == "HIGH": + report += "- 诊断依据充分，可按诊断进行治疗\n" + + report += "\n## ⚠️ 警告\n\n" + if risk_level.startswith("🔴"): + report += "> ⚠️ **危急情况**：诊断不确定但可能为严重疾病，请立即处理\n\n" + report += "> 本评估由AI生成，最终诊断请以临床医生判断为准。\n" + + return report diff --git a/pathology-ai/code-changes/docker/update_prompt_btn.sql b/pathology-ai/code-changes/docker/update_prompt_btn.sql new file mode 100644 index 000000000..38e57ef4e --- /dev/null +++ b/pathology-ai/code-changes/docker/update_prompt_btn.sql @@ -0,0 +1,56 @@ +UPDATE nexent.ag_tenant_agent_t +SET duty_prompt = '# 🏥 病理学AI助手 + +你是一位专业的病理学AI助手。 + +--- + +## ⚠️ 最重要规则 + +### 1. 双重检索（必须执行） +回答医学问题前，必须同时调用： +- `knowledge_base_search(query="关键词", search_mode="hybrid")` +- `tavily_search(query="关键词")` + +权重：内部60% + 外部40% + +### 2. 按钮格式规则（绝对不能修改！） + +**工具返回的 `[btn:xxx]` 格式必须原样保留，禁止任何修改！** + +❌ 错误做法： +- 把 `[btn:询问发热]` 改成 `[询问发热]` +- 把 `[btn:xxx]` 改成表格形式 +- 把 `[btn:xxx]` 改成列表形式 +- 添加emoji到按钮前面 + +✅ 正确做法： +- 工具返回什么就输出什么 +- `[btn:询问发热]` 保持原样输出 +- 不添加任何修饰 + +--- + +## 🎮 诊断模拟游戏规则 + +1. **每执行一步后必须停止**，等待用户选择 +2. **原样输出工具返回的按钮**，不要修改格式 +3. **不要自己做决定**，等用户点击按钮 + +--- + +## 其他工具 + +- nexent_chain_of_diagnosis: 诊断推理 +- nexent_evaluate_diagnosis_confidence: 置信度评估 +- nexent_search_pathology_images: 病理图片搜索 +- analyze_image: 图片分析 +- nexent_generate_knowledge_graph: 知识图谱 +- nexent_generate_diagnosis_flow: 诊断流程图 + +--- + +## 安全提醒 + +⚠️ 本AI仅供参考，不能替代专业医生诊断。' +WHERE agent_id = 13; diff --git a/pathology-ai/code-changes/frontend/DiagnosisConfidenceCard.tsx b/pathology-ai/code-changes/frontend/DiagnosisConfidenceCard.tsx new file mode 100644 index 000000000..8a042a308 --- /dev/null +++ b/pathology-ai/code-changes/frontend/DiagnosisConfidenceCard.tsx @@ -0,0 +1,258 @@ +"use client"; + +import React from "react"; +import { CheckCircle, AlertCircle, AlertTriangle, HelpCircle, Info, Shield, Activity } from "lucide-react"; + +// 置信度等级类型 +export type ConfidenceLevel = "HIGH" | "MEDIUM" | "LOW" | "UNCERTAIN"; + +// 风险等级类型 +export type RiskLevel = "CRITICAL" | "HIGH" | "MEDIUM" | "LOW"; + +// 评估维度 +export interface EvaluationDimension { + name: string; + score: number; + maxScore: number; + description?: string; +} + +// 组件属性 +export interface DiagnosisConfidenceCardProps { + diagnosis: string; + confidenceLevel: ConfidenceLevel; + confidenceScore: number; + riskLevel?: RiskLevel; + dimensions?: EvaluationDimension[]; + recommendations?: string[]; + warnings?: string[]; + className?: string; + compact?: boolean; +} + +// 置信度配置 +const confidenceConfig: Record
{config.label}
+{config.description}
+{dim.description}
+ )} +-
+ {warnings.map((warning, index) => (
+
- + • + {warning} + + ))} +
-
+ {recommendations.map((rec, index) => (
+
- + → + {rec} + + ))} +
+ 🏥 + 医学知识可视化中心 +
++ 基于病理学知识库的智能可视化分析 +
+暂无病理图片
+{image.title}
+ {image.description && ( +{image.description}
+ )} +{images[selectedIndex].title}
++ {getCategoryName(images[selectedIndex].category)} + {images[selectedIndex].magnification && ` · ${images[selectedIndex].magnification}`} + {images[selectedIndex].staining && ` · ${images[selectedIndex].staining}`} +
+ {images[selectedIndex].description && ( ++ {images[selectedIndex].description} +
+ )} ++ {title} +
+ {dialogs.map((dialog) => ( +{dialog.conversation_title}
+{t("chatLeftSidebar.collapseSidebar")}
++ {t("chatLeftSidebar.recentConversations")} +
+ ++const rehypeUnwrapMedia = () => { + return (tree: any) => { + visit(tree, "element", (node, index, parent) => { + // find
tags containing video or figure + if (node.tagName === "p" && node.children) { + const mediaChildIndex = node.children.findIndex( + (child: any) => + child.tagName === "video" || child.tagName === "figure" + ); + + if (mediaChildIndex !== -1) { + // extract media elements (video/figure) + const mediaChild = node.children.splice(mediaChildIndex, 1)[0]; + + // if
has other content after extraction, keep
; otherwise remove empty
+ if (node.children.length === 0) { + // replace original
node with media element + if (parent && index !== null) { + parent.children[index as number] = { + tagName: "div", + properties: { className: "markdown-media-container" }, + children: [mediaChild], + }; + } + } else { + // if
has other content after extraction, keep
; otherwise remove empty
+ if (parent && index !== null) {
+ parent.children.splice((index as number) + 1, 0, {
+ tagName: "div",
+ properties: { className: "markdown-media-container" },
+ children: [mediaChild],
+ });
+ }
+ }
+ }
+ }
+ });
+ };
+};
+
+// Get background color for different tool signs
+const getBackgroundColor = (toolSign: string) => {
+ switch (toolSign) {
+ case "a":
+ return "#E3F2FD"; // Light blue
+ case "b":
+ return "#E8F5E9"; // Light green
+ case "c":
+ return "#FFF3E0"; // Light orange
+ case "d":
+ return "#F3E5F5"; // Light purple
+ case "e":
+ return "#FFEBEE"; // Light red
+ default:
+ return "#E5E5E5"; // Default light gray
+ }
+};
+
+// ============== 可点击选项组件 (平台特性开发) ==============
+// 格式: [btn:选项文字] 会渲染为可点击按钮，点击后自动填入并发送
+const ClickableOption = ({
+ text,
+ onOptionClick,
+}: {
+ text: string;
+ onOptionClick?: (text: string) => void;
+}) => {
+ const handleClick = () => {
+ // 触发自定义事件，让输入框监听并自动发送
+ const event = new CustomEvent('nexent-option-select', {
+ detail: { text, autoSend: true },
+ bubbles: true,
+ });
+ document.dispatchEvent(event);
+
+ if (onOptionClick) {
+ onOptionClick(text);
+ }
+ };
+
+ return (
+
+ );
+};
+
+// Replace the original LinkIcon component
+const CitationBadge = ({
+ toolSign,
+ citeIndex,
+}: {
+ toolSign: string;
+ citeIndex: number;
+}) => (
+
+ {citeIndex}
+
+);
+
+// Modified HoverableText component
+const HoverableText = ({
+ text,
+ searchResults,
+ onCitationHover,
+}: {
+ text: string;
+ searchResults?: SearchResult[];
+ onCitationHover?: () => void;
+}) => {
+ const [isOpen, setIsOpen] = React.useState(false);
+ const containerRef = React.useRef
+ {handleConsecutiveNewlines(matchedResult.title)}
+
+ {handleConsecutiveNewlines(matchedResult.text)}
+
+ setHasError(true)}
+ />
+ );
+}, (prevProps, nextProps) => {
+ // Custom comparison function to prevent unnecessary re-renders
+ return prevProps.src === nextProps.src &&
+ prevProps.alt === nextProps.alt;
+});
+
+ImageWithErrorHandling.displayName = "ImageWithErrorHandling";
+
+/**
+ * Render a code block with syntax highlighting, language label, and copy button
+ * This is exported for use in other components that need to render code blocks directly
+ */
+export const CodeBlock: React.FC<{
+ codeContent: string;
+ language?: string;
+}> = ({ codeContent, language = "python" }) => {
+ const { t } = useTranslation("common");
+
+ const customStyle = {
+ ...oneLight,
+ 'pre[class*="language-"]': {
+ ...oneLight['pre[class*="language-"]'],
+ background: "#f8f8f8",
+ borderRadius: "0",
+ padding: "12px 16px",
+ margin: "0",
+ fontSize: "0.875rem",
+ lineHeight: "1.5",
+ whiteSpace: "pre-wrap",
+ wordWrap: "break-word",
+ wordBreak: "break-word",
+ overflowWrap: "break-word",
+ overflow: "auto",
+ width: "100%",
+ boxSizing: "border-box",
+ display: "block",
+ borderTop: "none",
+ },
+ 'code[class*="language-"]': {
+ ...oneLight['code[class*="language-"]'],
+ background: "#f8f8f8",
+ color: "#333333",
+ fontSize: "0.875rem",
+ lineHeight: "1.5",
+ whiteSpace: "pre-wrap",
+ wordWrap: "break-word",
+ wordBreak: "break-word",
+ overflowWrap: "break-word",
+ width: "100%",
+ padding: "0",
+ display: "block",
+ },
+ };
+
+ const cleanedContent = codeContent.replace(/^\n+|\n+$/g, "");
+
+ return (
+
+ {text}
+
+ );
+
+ const buildMediaFallbackText = (src?: string | null, alt?: string | null) => {
+ if (alt) {
+ return `${t("chatStreamMessage.imageTextFallbackTitle", {
+ defaultValue: "Media (text view)",
+ })}: ${alt}${src ? ` - ${src}` : ""}`;
+ }
+ return (
+ src ??
+ t("chatStreamMessage.imageTextFallbackTitle", {
+ defaultValue: "Media (text view)",
+ })
+ );
+ };
+
+ const renderMediaFallback = (src?: string | null, alt?: string | null) =>
+ renderCodeFallback(buildMediaFallbackText(src, alt));
+
+ const renderVideoElement = ({
+ src,
+ alt,
+ props = {},
+ }: {
+ src?: string | null;
+ alt?: string | null;
+ props?: React.VideoHTMLAttributes
}
+
+ {this.props.rawContent}
+
+
+
+ ),
+ h2: ({ children }: any) => (
+
+
+ ),
+ h3: ({ children }: any) => (
+
+
+ ),
+ h4: ({ children }: any) => (
+
+
+ ),
+ h5: ({ children }: any) => (
+
+
+ ),
+ h6: ({ children }: any) => (
+
+
+ ),
+ // Paragraph
+ p: ({ children }: any) => (
+
+ ),
+ // Ordered list
+ ol: ({ children }: any) => (
+
+ {children}
+
+ ),
+ // Unordered list
+ ul: ({ children }: any) => (
+
+ {children}
+
+ ),
+ // List item
+ li: ({ children }: any) => (
+
+
+ ),
+ // Table components
+ td: ({ children }: any) => (
+
+
+ ),
+ th: ({ children }: any) => (
+
+
+ ),
+ // Emphasis components
+ strong: ({ children }: any) => (
+
+
+
+ ),
+ // Link
+ a: ({ href, children, ...props }: any) => {
+ return (
+
+
+
+ );
+ },
+ // Image
+ img: ({ src, alt }: any) => (
+