AIclinicalresearch/docs/03-业务模块/ASL-AI智能文献/05-开发记录/2025-11-21-真实LLM集成完成报告.md

# 真实LLM集æˆ<C3A6>完æˆ<C3A6>报告

**日期**: 2025-11-21  
**任务**: å°†Mock AI替æ<C2BF>¢ä¸ºçœŸå®žLLM调用  
**状æ€?*: âœ?完æˆ<C3A6>  

---

## 📋 背景

### 之å‰<C3A5>的状æ€?
- �已完�Prompt 设计（v1.0.0-MVP�
- �已实�`llmScreeningService.ts`（真实LLM调用�
- âœ?已完æˆ<C3A6>测试框架和质é‡<C3A9>验è¯<C3A8>
- â<>?**问题**: `screeningService.ts` 中使ç”?`mockAIScreening` 生æˆ<C3A6>å<EFBFBD>‡æ•°æ<C2B0>?

### 用户需�
ä»?设置与å<C5BD>¯åŠ?页é<C2B5>¢ä¸Šä¼ çœŸå®žæ–‡çŒ®æ•°æ<C2B0>®å<C2AE>Žï¼Œ**使用真实çš?DeepSeek å’?Qwen API 进行筛é€?*，而ä¸<C3A4>是模拟数æ<C2B0>®ã€?

---

## âœ?完æˆ<C3A6>内容

### 1. 修改 `screeningService.ts`

**文件**: `backend/src/modules/asl/services/screeningService.ts`

#### 核心改动

**引入真实LLMæœ<C3A6>务**:
```typescript
import { llmScreeningService } from './llmScreeningService.js';
```

**替æ<C2BF>¢å¤„ç<E2809E>†é€»è¾‘**:
```typescript
// â<>?旧代ç <C3A7>（Mockï¼?
const result = await mockAIScreening(projectId, literature);

// âœ?新代ç <C3A7>（真实LLMï¼?
const screeningResult = await llmScreeningService.dualModelScreening(
  literature.id,
  literature.title,
  literature.abstract,
  picoCriteria,
  inclusionCriteria,
  exclusionCriteria,
  [models[0], models[1]],
  screeningConfig?.style || 'standard',
  literature.authors,
  literature.journal,
  literature.publicationYear
);
```

#### 新增功能

1. **从项目读å<C2BB>–PICOS标准**:
   ```typescript
   const project = await prisma.aslScreeningProject.findUnique({
     where: { id: projectId },
   });
   
   const picoCriteria = project.picoCriteria;
   const inclusionCriteria = project.inclusionCriteria;
   const exclusionCriteria = project.exclusionCriteria;
   ```

2. **支æŒ<C3A6>自定义模型选择**:
   ```typescript
   const models = screeningConfig?.models || ['deepseek-chat', 'qwen-max'];
   ```

3. **详细日志记录**:
   ```typescript
   logger.info('Processing literature', {
     literatureId: literature.id,
     title: literature.title?.substring(0, 50) + '...',
   });
   ```

4. **结果映射到数æ<C2B0>®åº“æ ¼å¼<C3A5>**:
   ```typescript
   const dbResult = {
     projectId,
     literatureId: literature.id,
     // DeepSeek结果
     dsModelName: screeningResult.deepseekModel,
     dsPJudgment: screeningResult.deepseek.judgment.P,
     // ... 完整的字段映�
   };
   ```

---

## 🔄 完整æµ<C3A6>程

### 用户æ“<C3A6>作æµ<C3A6>程
```
1. 访问"设置与å<C5BD>¯åŠ?页é<C2B5>¢
   �
2. 填写 PICOS 标准
   �
3. 上传 Excel 文献列表（例如：199篇）
   �
4. 点击"开始AIåˆ<C3A5>ç­›"
   �
5. å<>Žç«¯è‡ªåŠ¨å¤„ç<E2809E>†ï¼?
   a. 创建项目
   b. 导入文献
   c. å<>¯åŠ¨ç­›é€‰ä»»åŠ?
   �
6. 真实LLM处ç<E2809E>†ï¼ˆæ¯<C3A6>篇约10-15秒）
   a. 调用 DeepSeek API
   b. 调用 Qwen API
   c. 对比结果，检测冲�
   d. ä¿<C3A4>存到数æ<C2B0>®åº“
   �
7. å‰<C3A5>端自动跳转åˆ?审核工作å<C593>?
   �
8. 显示真实的AI筛选结�
```

### 技术æµ<C3A6>ç¨?

```
å‰<EFBFBD>端: TitleScreeningSettings.tsx
  �POST /api/v1/asl/literatures/import
  
å<EFBFBD>Žç«¯: literatureController.ts
  �importLiteratures()
  �startScreeningTask()
  
å<EFBFBD>Žç«¯: screeningService.ts
  �processLiteraturesInBackground()
  �for each literature:
      �llmScreeningService.dualModelScreening()
      
å<EFBFBD>Žç«¯: llmScreeningService.ts
  �Promise.all([
      screenWithModel('deepseek-chat', ...),
      screenWithModel('qwen-max', ...),
    ])
  
å<EFBFBD>Žç«¯: LLMFactory
  �getAdapter('deepseek-v3')
  �getAdapter('qwen3-72b')
  
真实API调用
  �DeepSeek API
  �Qwen API
  
结果ä¿<EFBFBD>å­˜
  �AslScreeningResult �
  
å‰<EFBFBD>端: ScreeningWorkbench.tsx
  �GET /api/v1/asl/projects/:projectId/screening-results
  �显示真实结果
```

---

## â<>±ï¸<C3AF> 性能预期

### å<>•ç¯‡æ–‡çŒ®å¤„ç<E2809E>†æ—¶é—´
| 步骤 | 耗时（串行） |
|-----|------------|
| DeepSeek API 调用 | 5-10�|
| Qwen API 调用 | 5-10�|
| 结果ä¿<C3A4>å­˜ | 0.1ç§?|
| **总计** | **10-20�* |

### 批é‡<C3A9>处ç<E2809E>†æ—¶é—´ï¼?99篇）
| 模å¼<C3A5> | 耗时 | 说明 |
|-----|------|-----|
| **串行处ç<E2809E>†** | 33-66分钟 | 当å‰<C3A5>实现（é<CB86>¿å…<C3A5>APIé™<C3A9>æµ<C3A6>）|
| 并å<C2B6>‘处ç<E2809E>†ï¼?个） | 11-22分钟 | å<>¯é€‰ä¼˜åŒ–（需测试ï¼?|
| 并å<C2B6>‘处ç<E2809E>†ï¼?0个） | 3-7分钟 | 风险：å<C5A1>¯èƒ½è§¦å<C2A6>‘APIé™<C3A9>é¢<C3A9> |

**当å‰<C3A5>ç­–ç•¥**: 串行处ç<E2809E>†ï¼ˆç¨³å®šä¼˜å…ˆï¼‰

---

## 🎯 与Mockæ•°æ<C2B0>®çš„对æ¯?

### Mock æ•°æ<C2B0>®ï¼ˆæ—§ï¼?
```javascript
// â<>?å<>‡æ•°æ<C2B0>?
dsPEvidence: "模拟è¯<C3A8>æ<EFBFBD>®: 研究人群与PICO中的P标准匹é…<C3A9>"
dsReason: "基于标题和摘è¦<C3A8>分æž<C3A6>，该文献符å<C2A6>ˆçº³å…¥æ ‡å‡†ã€?
dsConclusion: randomConclusion() // éš<C3A9>机ï¼?

// 特点�
- 1秒完�99�
- è¯<C3A8>æ<EFBFBD>®éƒ½æ˜¯"模拟è¯<C3A8>æ<EFBFBD>®"
- 判断结果éš<C3A9>机生æˆ<C3A6>
```

### 真实LLM（新�
```javascript
// âœ?真实数æ<C2B0>®
dsPEvidence: "This study included adult patients with type 2 diabetes mellitus aged 18 years or older, which matches the population criteria."
dsReason: "The study population consists of T2DM patients, the intervention is an SGLT2 inhibitor (empagliflozin), the comparator is placebo, and the study design is a randomized controlled trial. All PICO criteria are met. The study reports on cardiovascular outcomes including MACE, heart failure hospitalization, and cardiovascular death, which are the outcomes of interest."
dsConclusion: "include" // AI真实判断�

// 特点�
- 33-66分钟完æˆ<C3A6>199ç¯?
- è¯<C3A8>æ<EFBFBD>®å¼•ç”¨æ–‡çŒ®åŽŸæ–‡
- 判断基于Prompt v1.0.0-MVP
- 准确率：60%（首次测试）
```

---

## ðŸ”<C5B8> æ•°æ<C2B0>®éªŒè¯<C3A8>

### 验è¯<C3A8>方法
```bash
cd AIclinicalresearch/backend
node check-data.mjs
```

### 预期输出（真实数æ<C2B0>®ï¼‰
```
🔬 筛选结果样�
  [1] 文献: Assessment of Thrombectomy versus Combined...
      DeepSeek: include (P:match, I:partial, C:mismatch, S:match)
      Qwen:     exclude (P:mismatch, I:mismatch, C:partial, S:match)
      冲çª<C3A7>状æ€? conflict
      是å<C2AF>¦æœ‰è¯<C3A8>æ<EFBFBD>? DeepSeek=true, Qwen=true  âœ?

  è¯<C3A8>æ<EFBFBD>®ç¤ºä¾‹:
  - dsPEvidence: "The study population consists of..."
  - qwenPEvidence: "Patients with acute ischemic stroke..."
```

---

## 📊 è´¨é‡<C3A9>ä¿<C3A4>éšœ

### 已实现的质é‡<C3A9>措施

1. **JSON Schema 验è¯<C3A8>**:
   - 所有LLM输出必须通过Schema验è¯<C3A8>
   - ä¸<C3A4>å<EFBFBD>ˆæ ¼çš„输出会被拒ç»<C3A7>

2. **错误处ç<E2809E>†**:
   - å<>•ç¯‡æ–‡çŒ®å¤±è´¥ä¸<C3A4>å½±å“<C3A5>整体任åŠ?
   - 详细错误日志记录

3. **进度追踪**:
   - �0篇更新一次进�
   - 实时统计æˆ<C3A6>功/冲çª<C3A7>/失败æ•?

4. **å<>¯è¿½æº¯æ€?*:
   - 记录原始LLM输出（`rawOutput`�
   - 记录Prompt版本（`promptVersion`�
   - 记录处ç<E2809E>†æ—¶é—´ï¼ˆ`aiProcessedAt`ï¼?

---

## 🚀 测试步骤

### Step 1: 准备测试数æ<C2B0>®
```
使用现有测试文件:
- PICOS: docs/.../测试案例的PICOSã€<C3A3>纳入标准ã€<C3A3>排除标å‡?txt
- Excel: docs/.../Test Cases.xlsx (199篇文�
```

### Step 2: 执行测试
1. å<>¯åŠ¨å<C2A8>Žç«¯: `cd backend && npm run dev`
2. å<>¯åŠ¨å‰<C3A5>端: `cd frontend-v2 && npm run dev`
3. 访问: `http://localhost:3001`
4. 填写PICOS + 上传Excel
5. 点击"开始AIåˆ<C3A5>ç­›"
6. **等待30-60分钟**�99篇�0秒）
7. 查看审核工作å<C593>?

### Step 3: 验è¯<C3A8>结果
```bash
cd backend
node check-data.mjs
```

**检查项**:
- [ ] 所有文献都有筛选结�
- [ ] è¯<C3A8>æ<EFBFBD>®ä¸<C3A4>å†<C3A5>æ˜?模拟è¯<C3A8>æ<EFBFBD>®"
- [ ] è¯<C3A8>æ<EFBFBD>®åŒ…å<E280A6>«æ–‡çŒ®åŽŸæ–‡å¼•ç”¨
- [ ] 判断ç<C2AD>†ç”±è¯¦ç»†ä¸”符å<C2A6>ˆé€»è¾‘
- [ ] 冲çª<C3A7>检测准确（conclusionä¸<C3A4>å<EFBFBD>Œï¼?

---

## âš ï¸<C3AF> 注æ„<C3A6>事项

### API密钥é…<C3A9>ç½®
ç¡®ä¿<EFBFBD>环境å<EFBFBD>˜é‡<EFBFBD>å·²é…<EFBFBD>ç½?
```bash
# .env
DEEPSEEK_API_KEY=sk-xxxxx
QWEN_API_KEY=sk-xxxxx
```

### APIé™<C3A9>æµ<C3A6>
- DeepSeek: 60 RPM（æ¯<C3A6>分钟请求数）
- Qwen: 60 RPM

**当å‰<C3A5>ç­–ç•¥**: 串行处ç<E2809E>†ï¼Œä¸<C3A4>会触å<C2A6>‘é™<C3A9>æµ?

### æˆ<C3A6>本估算
- DeepSeek: ~$0.001/�× 199 = **$0.20**
- Qwen: ~$0.001/�× 199 = **$0.20**
- **总计**: **$0.40** / 次完整测�

---

## 💡 优化建议

### 短期优化（Week 2 - Day 4-5�
1. **并å<C2B6>‘控制**: 改为3个并å<C2B6>‘（33分钟 â†?11分钟ï¼?
2. **进度显示**: å‰<C3A5>端轮询显示进度百分æ¯?
3. **错误é‡<C3A9>试**: 失败的文献自动é‡<C3A9>è¯?æ¬?

### 中期优化（Week 3�
1. **消æ<CB86>¯é˜Ÿåˆ—**: 使用Bull Queue异步处ç<E2809E>†
2. **批é‡<C3A9>优化**: 使用批é‡<C3A9>API接å<C2A5>£ï¼ˆå¦‚果有ï¼?
3. **缓存机制**: 相å<C2B8>Œæ–‡çŒ®ä¸<C3A4>é‡<C3A9>å¤<C3A5>ç­›é€?

---

## ðŸ“<C5B8> 相关文件

### 修改的文�
- `backend/src/modules/asl/services/screeningService.ts` â­?

### ä¾<C3A4>赖的文件（已存在）
- `backend/src/modules/asl/services/llmScreeningService.ts`
- `backend/src/modules/asl/schemas/screening.schema.ts`
- `backend/prompts/asl/screening/v1.0.0-mvp.txt`
- `backend/src/common/llm/adapters/LLMFactory.ts`

### 测试文件
- `backend/scripts/test-llm-screening.ts`
- `backend/scripts/test-samples/asl-test-literatures.json`

---

## 🎉 æˆ<C3A6>果总结

### 已实�
âœ?真实LLM调用替æ<C2BF>¢Mockæ•°æ<C2B0>®  
âœ?从项目读å<C2BB>–PICOS标准  
âœ?å<>Œæ¨¡åž‹å¹¶è¡Œç­›é€? 
âœ?冲çª<C3A7>检测与标记  
�完整的日志追� 
âœ?错误处ç<E2809E>†æœºåˆ¶  

### 待优�
âš ï¸<EFBFBD> 处ç<E2809E>†æ—¶é—´è¾ƒé•¿ï¼?0-60分钟ï¼? 
âš ï¸<EFBFBD> 串行处ç<E2809E>†ï¼ˆå<CB86>¯æ”¹ä¸ºå¹¶å<C2B6>‘ï¼? 
âš ï¸<EFBFBD> å‰<C3A5>端进度显示（需优化轮询频率ï¼? 

---

## 🔗 å<>‚考文æ¡?

- [Prompt设计与测试完æˆ<EFBFBD>报告](./2025-11-18-Prompt设计与测试完æˆ<C3A6>报å‘?md)
- [å<EFBFBD>’中数æ<EFBFBD>®æ³›åŒ–测试报告](./2025-11-18-å<>’中数æ<C2B0>®æ³›åŒ–测试报告.md)
- [任务分解](../04-å¼€å<E282AC>‘计åˆ?03-任务分解.md)

---

**报告�*: AI Assistant  
**日期**: 2025-11-21  
**版本**: v1.0.0