feat(ssa): Complete QPER architecture - Query, Planner, Execute, Reflection layers
Implement the full QPER intelligent analysis pipeline: - Phase E+: Block-based standardization for all 7 R tools, DynamicReport renderer, Word export enhancement - Phase Q: LLM intent parsing with dynamic Zod validation against real column names, ClarificationCard component, DataProfile is_id_like tagging - Phase P: ConfigLoader with Zod schema validation and hot-reload API, DecisionTableService (4-dimension matching), FlowTemplateService with EPV protection, PlannedTrace audit output - Phase R: ReflectionService with statistical slot injection, sensitivity analysis conflict rules, ConclusionReport with section reveal animation, conclusion caching API, graceful R error classification End-to-end test: 40/40 passed across two complete analysis scenarios. Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
172
backend/scripts/seed-ssa-intent-prompt.ts
Normal file
172
backend/scripts/seed-ssa-intent-prompt.ts
Normal file
@@ -0,0 +1,172 @@
|
||||
/**
|
||||
* SSA Intent Prompt Seed 脚本
|
||||
*
|
||||
* 将 SSA_QUERY_INTENT prompt 写入 capability_schema.prompt_templates
|
||||
* 运行: npx tsx scripts/seed-ssa-intent-prompt.ts
|
||||
*/
|
||||
|
||||
import { PrismaClient, PromptStatus } from '@prisma/client';
|
||||
|
||||
const prisma = new PrismaClient();
|
||||
|
||||
const SSA_INTENT_PROMPT = `你是一个临床统计分析意图理解引擎。你的任务是根据用户的自然语言描述和数据画像，解析出结构化的分析意图。
|
||||
|
||||
## 输入信息
|
||||
|
||||
### 用户请求
|
||||
{{userQuery}}
|
||||
|
||||
### 数据画像
|
||||
{{dataProfile}}
|
||||
|
||||
### 可用统计工具
|
||||
{{availableTools}}
|
||||
|
||||
## 你的任务
|
||||
|
||||
请分析用户的请求，输出一个 JSON 对象（不要输出任何其他内容，只输出 JSON）：
|
||||
|
||||
\`\`\`json
|
||||
{
|
||||
"goal": "comparison | correlation | regression | descriptive | cohort_study",
|
||||
"outcome_var": "结局变量名（Y），必须是数据画像中存在的列名，如果无法确定则为 null",
|
||||
"outcome_type": "continuous | binary | categorical | ordinal | datetime | null",
|
||||
"predictor_vars": ["自变量名列表（X），必须是数据画像中存在的列名"],
|
||||
"predictor_types": ["对应每个自变量的类型"],
|
||||
"grouping_var": "分组变量名，必须是数据画像中存在的列名，如果无法确定则为 null",
|
||||
"design": "independent | paired | longitudinal | cross_sectional",
|
||||
"confidence": 0.0到1.0之间的数字,
|
||||
"reasoning": "你的推理过程，用1-2句话说明为什么这样解析"
|
||||
}
|
||||
\`\`\`
|
||||
|
||||
## 关键规则
|
||||
|
||||
1. **变量名必须精确匹配数据画像中的列名**，不要翻译、缩写或改写。如果数据里是 "Blood_Pressure"，你就输出 "Blood_Pressure"，不要输出 "BP"。
|
||||
2. 如果用户没有明确指出变量，请根据数据画像中的变量类型合理推断，但 confidence 应相应降低。
|
||||
3. goal 为 "descriptive" 时，不需要 outcome_var 和 predictor_vars。
|
||||
|
||||
## Confidence 评分准则（严格按此打分）
|
||||
|
||||
- **0.9 - 1.0**: 用户的原话中明确指定了结局变量(Y)和至少一个自变量(X)，且这些变量在数据画像中存在。
|
||||
- **0.7 - 0.8**: 用户指出了 Y 变量，但 X 需要根据数据类型推断；或用户的意图清晰但有轻微歧义。
|
||||
- **0.5 - 0.6**: 用户意图大致清楚（如"帮我比较一下"），但没有具体指出任何变量名。
|
||||
- **< 0.5**: 用户只说了"帮我分析一下"这样的模糊表达，既没有明确 Y 也没有明确 X，必须追问。
|
||||
|
||||
## Few-Shot 示例
|
||||
|
||||
### 示例 1：明确的差异比较
|
||||
用户: "帮我比较 Treatment 组和 Control 组的 SBP 有没有差异"
|
||||
数据画像中有: Group [categorical, 2个水平: Treatment/Control], SBP [numeric]
|
||||
输出:
|
||||
\`\`\`json
|
||||
{"goal":"comparison","outcome_var":"SBP","outcome_type":"continuous","predictor_vars":["Group"],"predictor_types":["binary"],"grouping_var":"Group","design":"independent","confidence":0.95,"reasoning":"用户明确指定了分组变量Group和结局变量SBP，要求比较两组差异"}
|
||||
\`\`\`
|
||||
|
||||
### 示例 2：相关分析
|
||||
用户: "年龄和血压有关系吗？"
|
||||
数据画像中有: Age [numeric], Blood_Pressure [numeric], Gender [categorical]
|
||||
输出:
|
||||
\`\`\`json
|
||||
{"goal":"correlation","outcome_var":"Blood_Pressure","outcome_type":"continuous","predictor_vars":["Age"],"predictor_types":["continuous"],"grouping_var":null,"design":"independent","confidence":0.85,"reasoning":"用户想了解Age和Blood_Pressure的关系，两者都是连续变量，适合相关分析"}
|
||||
\`\`\`
|
||||
|
||||
### 示例 3：多因素回归
|
||||
用户: "什么因素影响患者的死亡率？"
|
||||
数据画像中有: Death [categorical, 2个水平: 0/1], Age [numeric], BMI [numeric], Smoking [categorical, 2个水平: Yes/No], Stage [categorical, 4个水平]
|
||||
输出:
|
||||
\`\`\`json
|
||||
{"goal":"regression","outcome_var":"Death","outcome_type":"binary","predictor_vars":["Age","BMI","Smoking","Stage"],"predictor_types":["continuous","continuous","binary","categorical"],"grouping_var":null,"design":"independent","confidence":0.8,"reasoning":"用户想分析影响死亡率的因素，Death是二分类结局，其余变量作为预测因素纳入logistic回归"}
|
||||
\`\`\`
|
||||
|
||||
### 示例 4：模糊表达 — 需要追问
|
||||
用户: "帮我分析一下这份数据"
|
||||
数据画像中有: 10个变量
|
||||
输出:
|
||||
\`\`\`json
|
||||
{"goal":"descriptive","outcome_var":null,"outcome_type":null,"predictor_vars":[],"predictor_types":[],"grouping_var":null,"design":"independent","confidence":0.35,"reasoning":"用户没有指定任何分析目标和变量，只能先做描述性统计，建议追问具体分析目的"}
|
||||
\`\`\`
|
||||
|
||||
### 示例 5：队列研究
|
||||
用户: "我想做一个完整的队列研究分析，看看新药对预后的影响"
|
||||
数据画像中有: Drug [categorical, 2个水平], Outcome [categorical, 2个水平: 0/1], Age [numeric], Gender [categorical], BMI [numeric], Comorbidity [categorical]
|
||||
输出:
|
||||
\`\`\`json
|
||||
{"goal":"cohort_study","outcome_var":"Outcome","outcome_type":"binary","predictor_vars":["Drug","Age","Gender","BMI","Comorbidity"],"predictor_types":["binary","continuous","binary","continuous","categorical"],"grouping_var":"Drug","design":"independent","confidence":0.85,"reasoning":"用户明确要做队列研究分析，Drug是暴露因素/分组变量，Outcome是结局，其余为协变量"}
|
||||
\`\`\`
|
||||
|
||||
请只输出 JSON，不要输出其他内容。`;
|
||||
|
||||
async function main() {
|
||||
console.log('🚀 开始写入 SSA Intent Prompt...\n');
|
||||
|
||||
const existing = await prisma.prompt_templates.findUnique({
|
||||
where: { code: 'SSA_QUERY_INTENT' }
|
||||
});
|
||||
|
||||
if (existing) {
|
||||
console.log('⚠️ SSA_QUERY_INTENT 已存在 (id=%d)，创建新版本...', existing.id);
|
||||
|
||||
const latestVersion = await prisma.prompt_versions.findFirst({
|
||||
where: { template_id: existing.id },
|
||||
orderBy: { version: 'desc' }
|
||||
});
|
||||
|
||||
const newVersion = (latestVersion?.version ?? 0) + 1;
|
||||
|
||||
// 归档旧的 ACTIVE 版本
|
||||
await prisma.prompt_versions.updateMany({
|
||||
where: { template_id: existing.id, status: 'ACTIVE' },
|
||||
data: { status: 'ARCHIVED' }
|
||||
});
|
||||
|
||||
await prisma.prompt_versions.create({
|
||||
data: {
|
||||
template_id: existing.id,
|
||||
version: newVersion,
|
||||
content: SSA_INTENT_PROMPT,
|
||||
model_config: { model: 'deepseek-v3', temperature: 0.3, maxTokens: 2048 },
|
||||
status: 'ACTIVE',
|
||||
changelog: `Phase Q v1.0: 5 组 Few-Shot + Confidence Rubric 客观化`,
|
||||
created_by: 'system-seed',
|
||||
}
|
||||
});
|
||||
|
||||
console.log(' ✅ 新版本 v%d 已创建并设为 ACTIVE', newVersion);
|
||||
} else {
|
||||
console.log('📝 创建 SSA_QUERY_INTENT 模板...');
|
||||
|
||||
const template = await prisma.prompt_templates.create({
|
||||
data: {
|
||||
code: 'SSA_QUERY_INTENT',
|
||||
name: 'SSA 意图理解 Prompt',
|
||||
module: 'SSA',
|
||||
description: 'Phase Q — 将用户自然语言转化为结构化的统计分析意图 (ParsedQuery)',
|
||||
variables: ['userQuery', 'dataProfile', 'availableTools'],
|
||||
}
|
||||
});
|
||||
|
||||
await prisma.prompt_versions.create({
|
||||
data: {
|
||||
template_id: template.id,
|
||||
version: 1,
|
||||
content: SSA_INTENT_PROMPT,
|
||||
model_config: { model: 'deepseek-v3', temperature: 0.3, maxTokens: 2048 },
|
||||
status: 'ACTIVE',
|
||||
changelog: 'Phase Q v1.0: 初始版本，5 组 Few-Shot + Confidence Rubric',
|
||||
created_by: 'system-seed',
|
||||
}
|
||||
});
|
||||
|
||||
console.log(' ✅ 模板 id=%d + 版本 v1 已创建', template.id);
|
||||
}
|
||||
|
||||
console.log('\n✅ SSA Intent Prompt 写入完成！');
|
||||
}
|
||||
|
||||
main()
|
||||
.catch(e => {
|
||||
console.error('❌ 写入失败:', e);
|
||||
process.exit(1);
|
||||
})
|
||||
.finally(() => prisma.$disconnect());
|
||||
174
backend/scripts/seed-ssa-reflection-prompt.ts
Normal file
174
backend/scripts/seed-ssa-reflection-prompt.ts
Normal file
@@ -0,0 +1,174 @@
|
||||
/**
|
||||
* SSA Reflection Prompt Seed 脚本
|
||||
*
|
||||
* 将 SSA_REFLECTION prompt 写入 capability_schema.prompt_templates
|
||||
* 运行: npx tsx scripts/seed-ssa-reflection-prompt.ts
|
||||
*
|
||||
* Phase R — 论文级结论生成 Prompt
|
||||
* 特性：
|
||||
* - 统计量槽位注入（反幻觉）
|
||||
* - 敏感性分析冲突处理准则
|
||||
* - 6 要素结构化 JSON 输出
|
||||
* - 基于 decision_trace 的方法学说明
|
||||
*/
|
||||
|
||||
import { PrismaClient } from '@prisma/client';
|
||||
|
||||
const prisma = new PrismaClient();
|
||||
|
||||
const SSA_REFLECTION_PROMPT = `你是一位高级生物统计师，请基于以下分析结果生成论文级结论。
|
||||
|
||||
## 分析背景
|
||||
|
||||
分析目标：{{goal}}
|
||||
分析标题：{{title}}
|
||||
采用方法：{{methodology}}
|
||||
样本信息：{{sampleInfo}}
|
||||
|
||||
## 方法选择的决策轨迹
|
||||
|
||||
请据此撰写方法学说明，不得臆造选择理由：
|
||||
|
||||
匹配规则：{{decision_trace.matched_rule}}
|
||||
主要工具：{{decision_trace.primary_tool}}
|
||||
{{#if decision_trace.fallback_tool}}备选工具：{{decision_trace.fallback_tool}}（触发条件：{{decision_trace.switch_condition}}）{{/if}}
|
||||
决策推理：{{decision_trace.reasoning}}
|
||||
{{#if decision_trace.epv_warning}}⚠️ EPV 警告：{{decision_trace.epv_warning}}{{/if}}
|
||||
|
||||
## 各步骤统计结果
|
||||
|
||||
⚠️ 以下数值由系统自动注入，你必须原样引用，不得修改、四舍五入或重新表述任何数字。
|
||||
|
||||
{{#each findings}}
|
||||
### 步骤 {{step_number}}：{{tool_name}}（{{tool_code}}）
|
||||
- 使用方法：{{method}}
|
||||
{{#if statistic}}- 统计量（{{statistic_name}}）：{{statistic}}{{/if}}
|
||||
{{#if p_value}}- P 值：{{p_value}}{{/if}}
|
||||
{{#if effect_size}}- 效应量（{{effect_size_name}}）：{{effect_size}}{{/if}}
|
||||
{{#if ci_lower}}- 95% 置信区间：{{ci_lower}} ~ {{ci_upper}}{{/if}}
|
||||
- 显著性：{{#if is_significant}}显著（P < 0.05）{{else}}不显著（P ≥ 0.05）{{/if}}
|
||||
{{#if group_stats}}
|
||||
- 各组统计：
|
||||
{{#each group_stats}} · {{group}} 组：n={{n}}{{#if mean}}, 均值={{mean}}, SD={{sd}}{{/if}}{{#if median}}, 中位数={{median}}{{/if}}
|
||||
{{/each}}
|
||||
{{/if}}
|
||||
|
||||
{{/each}}
|
||||
|
||||
## 冲突处理准则（强制执行）
|
||||
|
||||
当主分析与敏感性分析的显著性结论不一致时：
|
||||
1. 在 limitations 数组中必须包含："敏感性分析未得到一致结论，结果的稳健性（Robustness）较弱，需谨慎解释临床意义"
|
||||
2. 在 key_findings 中以主分析结果为基准报告，但需加注"（注：敏感性分析未验证此结论）"
|
||||
3. 严禁选择性报告、强行拼凑显著性
|
||||
4. 当所有分析方向一致时，在 key_findings 中可强调"敏感性分析进一步验证了结论的稳健性"
|
||||
|
||||
## 输出要求
|
||||
|
||||
请输出一个严格的 JSON 对象（不要输出任何其他内容，只输出 JSON），包含以下字段：
|
||||
|
||||
\`\`\`json
|
||||
{
|
||||
"executive_summary": "200-500字的综合摘要，涵盖研究目的、主要发现和临床意义。使用论文'结果'章节的行文风格，可直接复制到论文中。引用统计量时必须与上方数值完全一致。",
|
||||
"key_findings": [
|
||||
"发现1：具体描述，含统计量和P值（必须与上方数值一致）",
|
||||
"发现2：..."
|
||||
],
|
||||
"statistical_summary": {
|
||||
"total_tests": 3,
|
||||
"significant_results": 2,
|
||||
"methods_used": ["独立样本T检验", "Mann-Whitney U检验"]
|
||||
},
|
||||
"methodology": "方法学说明段落：基于上方决策轨迹撰写，解释为什么选择此方法、是否发生降级、数据满足了哪些假设。使用论文'方法'章节的行文风格。",
|
||||
"limitations": [
|
||||
"局限性1",
|
||||
"局限性2"
|
||||
],
|
||||
"recommendations": [
|
||||
"建议1：基于分析结果的后续研究建议",
|
||||
"建议2：..."
|
||||
]
|
||||
}
|
||||
\`\`\`
|
||||
|
||||
## 关键规则
|
||||
|
||||
1. **所有数值必须与上方统计结果完全一致**，不得修改、四舍五入、近似或重新表述
|
||||
2. executive_summary 必须是完整的论文段落，不是要点列表
|
||||
3. key_findings 每条必须包含具体的统计量和 P 值
|
||||
4. methodology 必须基于决策轨迹撰写，说明方法选择的理由
|
||||
5. limitations 至少包含 1 条，如果数据有局限则如实报告
|
||||
6. 请只输出 JSON，不要输出其他内容`;
|
||||
|
||||
async function main() {
|
||||
console.log('🚀 开始写入 SSA Reflection Prompt...\n');
|
||||
|
||||
const existing = await prisma.prompt_templates.findUnique({
|
||||
where: { code: 'SSA_REFLECTION' }
|
||||
});
|
||||
|
||||
if (existing) {
|
||||
console.log('⚠️ SSA_REFLECTION 已存在 (id=%d)，创建新版本...', existing.id);
|
||||
|
||||
const latestVersion = await prisma.prompt_versions.findFirst({
|
||||
where: { template_id: existing.id },
|
||||
orderBy: { version: 'desc' }
|
||||
});
|
||||
|
||||
const newVersion = (latestVersion?.version ?? 0) + 1;
|
||||
|
||||
await prisma.prompt_versions.updateMany({
|
||||
where: { template_id: existing.id, status: 'ACTIVE' },
|
||||
data: { status: 'ARCHIVED' }
|
||||
});
|
||||
|
||||
await prisma.prompt_versions.create({
|
||||
data: {
|
||||
template_id: existing.id,
|
||||
version: newVersion,
|
||||
content: SSA_REFLECTION_PROMPT,
|
||||
model_config: { model: 'deepseek-v3', temperature: 0.3, maxTokens: 4096 },
|
||||
status: 'ACTIVE',
|
||||
changelog: `Phase R v1.0: 6要素结构化JSON + 槽位注入 + 冲突处理准则`,
|
||||
created_by: 'system-seed',
|
||||
}
|
||||
});
|
||||
|
||||
console.log(' ✅ 新版本 v%d 已创建并设为 ACTIVE', newVersion);
|
||||
} else {
|
||||
console.log('📝 创建 SSA_REFLECTION 模板...');
|
||||
|
||||
const template = await prisma.prompt_templates.create({
|
||||
data: {
|
||||
code: 'SSA_REFLECTION',
|
||||
name: 'SSA 论文级结论生成 Prompt',
|
||||
module: 'SSA',
|
||||
description: 'Phase R — 将 StepResult[] 转化为论文级结论（6要素结构化JSON，含槽位注入反幻觉 + 敏感性冲突准则）',
|
||||
variables: ['goal', 'title', 'methodology', 'sampleInfo', 'decision_trace', 'findings'],
|
||||
}
|
||||
});
|
||||
|
||||
await prisma.prompt_versions.create({
|
||||
data: {
|
||||
template_id: template.id,
|
||||
version: 1,
|
||||
content: SSA_REFLECTION_PROMPT,
|
||||
model_config: { model: 'deepseek-v3', temperature: 0.3, maxTokens: 4096 },
|
||||
status: 'ACTIVE',
|
||||
changelog: 'Phase R v1.0: 初始版本，6要素JSON + 槽位注入 + 冲突处理准则',
|
||||
created_by: 'system-seed',
|
||||
}
|
||||
});
|
||||
|
||||
console.log(' ✅ 模板 id=%d + 版本 v1 已创建', template.id);
|
||||
}
|
||||
|
||||
console.log('\n✅ SSA Reflection Prompt 写入完成！');
|
||||
}
|
||||
|
||||
main()
|
||||
.catch(e => {
|
||||
console.error('❌ 写入失败:', e);
|
||||
process.exit(1);
|
||||
})
|
||||
.finally(() => prisma.$disconnect());
|
||||
494
backend/scripts/test-ssa-phase-q-e2e.ts
Normal file
494
backend/scripts/test-ssa-phase-q-e2e.ts
Normal file
@@ -0,0 +1,494 @@
|
||||
/**
|
||||
* SSA Phase Q — 端到端集成测试
|
||||
*
|
||||
* 完整链路：登录 → 创建会话+上传文件 → 数据画像 → LLM 意图解析 → 追问 → Q→P 规划
|
||||
*
|
||||
* 依赖：Node.js 后端 + PostgreSQL + Python extraction_service + LLM 服务
|
||||
* 运行方式：npx tsx scripts/test-ssa-phase-q-e2e.ts
|
||||
*
|
||||
* 测试数据：docs/03-业务模块/SSA-智能统计分析/05-测试文档/test.csv
|
||||
* 测试用户：13800000001 / 123456
|
||||
*/
|
||||
|
||||
import { readFileSync } from 'fs';
|
||||
import { join, dirname } from 'path';
|
||||
import { fileURLToPath } from 'url';
|
||||
|
||||
const __filename = fileURLToPath(import.meta.url);
|
||||
const __dirname = dirname(__filename);
|
||||
|
||||
const BASE_URL = 'http://localhost:3000';
|
||||
const TEST_PHONE = '13800000001';
|
||||
const TEST_PASSWORD = '123456';
|
||||
const TEST_CSV_PATH = join(__dirname, '../../docs/03-业务模块/SSA-智能统计分析/05-测试文档/test.csv');
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 工具函数
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
let passed = 0;
|
||||
let failed = 0;
|
||||
let skipped = 0;
|
||||
let token = '';
|
||||
let sessionId = '';
|
||||
|
||||
function assert(condition: boolean, testName: string, detail?: string) {
|
||||
if (condition) {
|
||||
console.log(` ✅ ${testName}`);
|
||||
passed++;
|
||||
} else {
|
||||
console.log(` ❌ ${testName}${detail ? ` — ${detail}` : ''}`);
|
||||
failed++;
|
||||
}
|
||||
}
|
||||
|
||||
function skip(testName: string, reason: string) {
|
||||
console.log(` ⏭️ ${testName} — 跳过：${reason}`);
|
||||
skipped++;
|
||||
}
|
||||
|
||||
function section(title: string) {
|
||||
console.log(`\n${'─'.repeat(60)}`);
|
||||
console.log(`📋 ${title}`);
|
||||
console.log('─'.repeat(60));
|
||||
}
|
||||
|
||||
function authHeaders(contentType?: string): Record<string, string> {
|
||||
const headers: Record<string, string> = {
|
||||
'Authorization': `Bearer ${token}`,
|
||||
};
|
||||
if (contentType) {
|
||||
headers['Content-Type'] = contentType;
|
||||
}
|
||||
return headers;
|
||||
}
|
||||
|
||||
async function apiPost(path: string, body: any, headers?: Record<string, string>): Promise<any> {
|
||||
const res = await fetch(`${BASE_URL}${path}`, {
|
||||
method: 'POST',
|
||||
headers: headers || authHeaders('application/json'),
|
||||
body: typeof body === 'string' ? body : JSON.stringify(body),
|
||||
});
|
||||
const text = await res.text();
|
||||
try {
|
||||
return { status: res.status, data: JSON.parse(text) };
|
||||
} catch {
|
||||
return { status: res.status, data: text };
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 1: 登录获取 Token
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testLogin() {
|
||||
section('测试 1: 登录认证');
|
||||
|
||||
try {
|
||||
const res = await apiPost('/api/v1/auth/login/password', {
|
||||
phone: TEST_PHONE,
|
||||
password: TEST_PASSWORD,
|
||||
}, { 'Content-Type': 'application/json' });
|
||||
|
||||
assert(res.status === 200, `登录返回 200（实际 ${res.status}）`, JSON.stringify(res.data).substring(0, 200));
|
||||
|
||||
if (res.status === 200 && res.data) {
|
||||
token = res.data?.data?.tokens?.accessToken || res.data?.accessToken || res.data?.token || '';
|
||||
assert(token.length > 0, '获取到 JWT Token', `token 长度: ${token.length}`);
|
||||
|
||||
if (res.data?.data?.user) {
|
||||
const user = res.data.data.user;
|
||||
console.log(` 用户信息: ${user.name || user.phone || 'N/A'}, 角色: ${user.role}`);
|
||||
}
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, '登录请求失败', e.message);
|
||||
}
|
||||
|
||||
if (!token) {
|
||||
console.log('\n ⚠️ Token 获取失败，后续测试无法继续');
|
||||
return false;
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 2: 创建会话 + 上传 test.csv
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testCreateSessionWithUpload() {
|
||||
section('测试 2: 创建会话 + 上传 test.csv');
|
||||
|
||||
try {
|
||||
const csvBuffer = readFileSync(TEST_CSV_PATH);
|
||||
assert(csvBuffer.length > 0, `test.csv 文件读取成功（${csvBuffer.length} bytes）`);
|
||||
|
||||
// 构建 multipart/form-data
|
||||
const formData = new FormData();
|
||||
const blob = new Blob([csvBuffer], { type: 'text/csv' });
|
||||
formData.append('file', blob, 'test.csv');
|
||||
|
||||
const res = await fetch(`${BASE_URL}/api/v1/ssa/sessions/`, {
|
||||
method: 'POST',
|
||||
headers: { 'Authorization': `Bearer ${token}` },
|
||||
body: formData,
|
||||
});
|
||||
|
||||
const data = await res.json();
|
||||
assert(res.status === 200, `创建会话返回 200（实际 ${res.status}）`, JSON.stringify(data).substring(0, 300));
|
||||
|
||||
if (data.sessionId) {
|
||||
sessionId = data.sessionId;
|
||||
assert(true, `会话 ID: ${sessionId}`);
|
||||
} else {
|
||||
assert(false, '未返回 sessionId', JSON.stringify(data).substring(0, 200));
|
||||
}
|
||||
|
||||
if (data.schema) {
|
||||
const schema = data.schema;
|
||||
assert(schema.columns?.length > 0, `数据 Schema 解析成功（${schema.columns?.length} 列）`);
|
||||
assert(schema.rowCount > 0, `行数: ${schema.rowCount}`);
|
||||
console.log(` 列名: ${schema.columns?.slice(0, 8).map((c: any) => c.name).join(', ')}...`);
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, '创建会话失败', e.message);
|
||||
}
|
||||
|
||||
if (!sessionId) {
|
||||
console.log('\n ⚠️ SessionId 获取失败，后续测试无法继续');
|
||||
return false;
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 3: 数据画像（Python DataProfiler）
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testDataProfile() {
|
||||
section('测试 3: 数据画像（Python DataProfiler）');
|
||||
|
||||
try {
|
||||
const res = await apiPost('/api/v1/ssa/workflow/profile', { sessionId });
|
||||
|
||||
assert(res.status === 200, `画像请求返回 200（实际 ${res.status}）`);
|
||||
|
||||
if (res.data?.success) {
|
||||
const profile = res.data.profile;
|
||||
assert(!!profile, '画像数据非空');
|
||||
|
||||
if (profile) {
|
||||
assert(profile.row_count > 0 || profile.totalRows > 0,
|
||||
`行数: ${profile.row_count || profile.totalRows}`);
|
||||
assert(profile.column_count > 0 || profile.totalColumns > 0,
|
||||
`列数: ${profile.column_count || profile.totalColumns}`);
|
||||
|
||||
const cols = profile.columns || [];
|
||||
if (cols.length > 0) {
|
||||
console.log(` 前 5 列类型:`);
|
||||
cols.slice(0, 5).forEach((c: any) => {
|
||||
console.log(` ${c.name || c.column_name}: ${c.type || c.dtype} (missing: ${c.missing_ratio ?? c.missingPercent ?? 'N/A'})`);
|
||||
});
|
||||
|
||||
// 检查 is_id_like 标记（Phase Q 防御性优化）
|
||||
const idLikeCols = cols.filter((c: any) => c.is_id_like === true);
|
||||
if (idLikeCols.length > 0) {
|
||||
assert(true, `检测到 ${idLikeCols.length} 个 ID-like 列: ${idLikeCols.map((c: any) => c.name || c.column_name).join(', ')}`);
|
||||
} else {
|
||||
console.log(' ℹ️ 未检测到 ID-like 列（test.csv 无 ID 列，符合预期）');
|
||||
}
|
||||
}
|
||||
}
|
||||
} else {
|
||||
assert(false, '画像生成失败', res.data?.error || JSON.stringify(res.data).substring(0, 200));
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, '画像请求异常', e.message);
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 4: LLM 意图解析（Phase Q 核心）
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testIntentParsing() {
|
||||
section('测试 4: LLM 意图理解（Phase Q 核心）');
|
||||
|
||||
const testQueries = [
|
||||
{
|
||||
name: '场景 A — 明确的差异比较',
|
||||
query: '比较 sex 不同组的 Yqol 有没有差别',
|
||||
expectGoal: 'comparison',
|
||||
expectHighConfidence: true,
|
||||
},
|
||||
{
|
||||
name: '场景 B — 相关分析',
|
||||
query: '分析 age 和 bmi 的相关性',
|
||||
expectGoal: 'correlation',
|
||||
expectHighConfidence: true,
|
||||
},
|
||||
{
|
||||
name: '场景 C — 回归分析',
|
||||
query: 'age、smoke、bmi 对 Yqol 的影响，做个多因素分析',
|
||||
expectGoal: 'regression',
|
||||
expectHighConfidence: true,
|
||||
},
|
||||
{
|
||||
name: '场景 D — 模糊意图（应触发追问）',
|
||||
query: '帮我分析一下这个数据',
|
||||
expectGoal: null, // 不确定
|
||||
expectHighConfidence: false,
|
||||
},
|
||||
{
|
||||
name: '场景 E — 描述统计',
|
||||
query: '描述一下数据的基本情况',
|
||||
expectGoal: 'descriptive',
|
||||
expectHighConfidence: true,
|
||||
},
|
||||
];
|
||||
|
||||
for (const tc of testQueries) {
|
||||
console.log(`\n 🔬 ${tc.name}`);
|
||||
console.log(` Query: "${tc.query}"`);
|
||||
|
||||
try {
|
||||
const startTime = Date.now();
|
||||
const res = await apiPost('/api/v1/ssa/workflow/intent', {
|
||||
sessionId,
|
||||
userQuery: tc.query,
|
||||
});
|
||||
const elapsed = Date.now() - startTime;
|
||||
|
||||
assert(res.status === 200, ` 返回 200（实际 ${res.status}）`, res.data?.error);
|
||||
|
||||
if (res.data?.success) {
|
||||
const intent = res.data.intent;
|
||||
console.log(` 耗时: ${elapsed}ms`);
|
||||
console.log(` Goal: ${intent?.goal}, Confidence: ${intent?.confidence}`);
|
||||
console.log(` Y: ${intent?.outcome_var || 'null'}, X: ${JSON.stringify(intent?.predictor_vars || [])}`);
|
||||
console.log(` Design: ${intent?.design}, needsClarification: ${intent?.needsClarification}`);
|
||||
|
||||
if (intent) {
|
||||
// 检查 goal 是否符合预期
|
||||
if (tc.expectGoal) {
|
||||
assert(intent.goal === tc.expectGoal,
|
||||
` Goal = ${tc.expectGoal}（实际 ${intent.goal}）`);
|
||||
}
|
||||
|
||||
// 检查置信度
|
||||
if (tc.expectHighConfidence) {
|
||||
assert(intent.confidence >= 0.7,
|
||||
` 高置信度 >= 0.7（实际 ${intent.confidence}）`);
|
||||
assert(!intent.needsClarification,
|
||||
` 无需追问（实际 needsClarification=${intent.needsClarification}）`);
|
||||
} else {
|
||||
// 模糊意图应该低置信度或触发追问
|
||||
const isLowConfOrClarify = intent.confidence < 0.7 || intent.needsClarification;
|
||||
assert(isLowConfOrClarify,
|
||||
` 低置信度或需追问（confidence=${intent.confidence}, needsClarification=${intent.needsClarification}）`);
|
||||
}
|
||||
|
||||
// 检查变量名是否来自真实数据（防幻觉校验）
|
||||
const realColumns = ['sex', 'smoke', 'age', 'bmi', 'mouth_open', 'bucal_relax',
|
||||
'toot_morph', 'root_number', 'root_curve', 'lenspace', 'denseratio',
|
||||
'Pglevel', 'Pgverti', 'Winter', 'presyp', 'flap', 'operation',
|
||||
'time', 'surgage', 'Yqol', 'times'];
|
||||
const realColumnsLower = realColumns.map(c => c.toLowerCase());
|
||||
|
||||
if (intent.outcome_var) {
|
||||
const isReal = realColumnsLower.includes(intent.outcome_var.toLowerCase());
|
||||
assert(isReal,
|
||||
` Y 变量 "${intent.outcome_var}" 存在于数据中`,
|
||||
`变量 "${intent.outcome_var}" 不在数据列名中（可能是 LLM 幻觉）`);
|
||||
}
|
||||
|
||||
if (intent.predictor_vars?.length > 0) {
|
||||
const allReal = intent.predictor_vars.every(
|
||||
(v: string) => realColumnsLower.includes(v.toLowerCase())
|
||||
);
|
||||
assert(allReal,
|
||||
` X 变量 ${JSON.stringify(intent.predictor_vars)} 全部存在于数据中`,
|
||||
`部分变量可能为 LLM 幻觉`);
|
||||
}
|
||||
|
||||
// 检查追问卡片（模糊意图时）
|
||||
if (intent.needsClarification && res.data.clarificationCards?.length > 0) {
|
||||
const cards = res.data.clarificationCards;
|
||||
console.log(` 追问卡片: ${cards.length} 张`);
|
||||
cards.forEach((card: any, i: number) => {
|
||||
console.log(` 卡片 ${i + 1}: ${card.question}`);
|
||||
card.options?.slice(0, 3).forEach((opt: any) => {
|
||||
console.log(` - ${opt.label}`);
|
||||
});
|
||||
});
|
||||
assert(cards[0].options?.length >= 2, ` 追问卡片有 >= 2 个选项`);
|
||||
}
|
||||
}
|
||||
} else {
|
||||
assert(false, ` Intent 解析失败`, res.data?.error || JSON.stringify(res.data).substring(0, 200));
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, ` ${tc.name} 请求异常`, e.message);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 5: Q→P 全链路（Intent → Plan）
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testQtoPPipeline() {
|
||||
section('测试 5: Q→P 全链路（Intent → WorkflowPlan）');
|
||||
|
||||
const testCases = [
|
||||
{
|
||||
name: '差异比较 → T 检验流程',
|
||||
query: '比较 sex 不同组的 Yqol 有没有差别',
|
||||
expectSteps: 2, // 描述统计 + 主分析（至少）
|
||||
expectTool: 'ST_',
|
||||
},
|
||||
{
|
||||
name: '回归分析 → Logistic 流程',
|
||||
query: 'age、smoke、bmi 对 Yqol 的预测作用，做个 Logistic 回归',
|
||||
expectSteps: 2,
|
||||
expectTool: 'ST_LOGISTIC',
|
||||
},
|
||||
];
|
||||
|
||||
for (const tc of testCases) {
|
||||
console.log(`\n 🔬 ${tc.name}`);
|
||||
console.log(` Query: "${tc.query}"`);
|
||||
|
||||
try {
|
||||
const startTime = Date.now();
|
||||
const res = await apiPost('/api/v1/ssa/workflow/plan', {
|
||||
sessionId,
|
||||
userQuery: tc.query,
|
||||
});
|
||||
const elapsed = Date.now() - startTime;
|
||||
|
||||
assert(res.status === 200, ` 返回 200（实际 ${res.status}）`, res.data?.error);
|
||||
|
||||
if (res.data?.success && res.data.plan) {
|
||||
const plan = res.data.plan;
|
||||
console.log(` 耗时: ${elapsed}ms`);
|
||||
console.log(` 标题: ${plan.title}`);
|
||||
console.log(` 步骤数: ${plan.total_steps}`);
|
||||
|
||||
assert(plan.total_steps >= tc.expectSteps,
|
||||
` 步骤数 >= ${tc.expectSteps}（实际 ${plan.total_steps}）`);
|
||||
|
||||
// 打印每步信息
|
||||
plan.steps?.forEach((step: any, i: number) => {
|
||||
const sensitivity = step.is_sensitivity ? ' [敏感性分析]' : '';
|
||||
const guardrail = step.switch_condition ? ` 🛡️${step.switch_condition}` : '';
|
||||
console.log(` 步骤 ${i + 1}: ${step.tool_name} (${step.tool_code})${sensitivity}${guardrail}`);
|
||||
});
|
||||
|
||||
// 检查是否包含期望的工具
|
||||
const hasExpectedTool = plan.steps?.some(
|
||||
(s: any) => s.tool_code?.startsWith(tc.expectTool)
|
||||
);
|
||||
assert(hasExpectedTool,
|
||||
` 包含 ${tc.expectTool}* 工具`,
|
||||
`工具列表: ${plan.steps?.map((s: any) => s.tool_code).join(', ')}`);
|
||||
|
||||
// 检查 PlannedTrace
|
||||
if (plan.planned_trace) {
|
||||
const trace = plan.planned_trace;
|
||||
console.log(` PlannedTrace:`);
|
||||
console.log(` Primary: ${trace.primaryTool}`);
|
||||
console.log(` Fallback: ${trace.fallbackTool || 'null'}`);
|
||||
console.log(` SwitchCondition: ${trace.switchCondition || 'null'}`);
|
||||
console.log(` Template: ${trace.templateUsed}`);
|
||||
assert(!!trace.primaryTool, ` PlannedTrace 包含 primaryTool`);
|
||||
assert(!!trace.templateUsed, ` PlannedTrace 包含 templateUsed`);
|
||||
} else {
|
||||
skip('PlannedTrace 检查', '计划中未返回 planned_trace');
|
||||
}
|
||||
|
||||
// EPV 警告检查
|
||||
if (plan.epv_warning) {
|
||||
console.log(` ⚠️ EPV Warning: ${plan.epv_warning}`);
|
||||
}
|
||||
|
||||
// 描述文字检查
|
||||
if (plan.description) {
|
||||
assert(plan.description.length > 10, ` 规划描述非空（${plan.description.length} 字符）`);
|
||||
console.log(` 描述: ${plan.description.substring(0, 100)}...`);
|
||||
}
|
||||
} else {
|
||||
assert(false, ` 规划失败`, res.data?.error || JSON.stringify(res.data).substring(0, 200));
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, ` ${tc.name} 请求异常`, e.message);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 运行所有测试
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function main() {
|
||||
console.log('\n🧪 SSA Phase Q+P — 端到端集成测试\n');
|
||||
console.log('测试链路：登录 → 上传 CSV → 数据画像 → LLM Intent → Q→P Plan');
|
||||
console.log(`测试用户：${TEST_PHONE}`);
|
||||
console.log(`后端地址：${BASE_URL}`);
|
||||
console.log(`测试文件：${TEST_CSV_PATH}\n`);
|
||||
|
||||
// 前置检查
|
||||
try {
|
||||
readFileSync(TEST_CSV_PATH);
|
||||
} catch {
|
||||
console.error('❌ test.csv 文件不存在，请检查路径');
|
||||
process.exit(1);
|
||||
}
|
||||
|
||||
try {
|
||||
const healthCheck = await fetch(`${BASE_URL}/health`).catch(() => null);
|
||||
if (!healthCheck || healthCheck.status !== 200) {
|
||||
console.error('❌ 后端服务未启动或不可达');
|
||||
process.exit(1);
|
||||
}
|
||||
console.log('✅ 后端服务可达\n');
|
||||
} catch {
|
||||
console.error('❌ 后端服务未启动或不可达');
|
||||
process.exit(1);
|
||||
}
|
||||
|
||||
// 顺序执行测试
|
||||
const loginOk = await testLogin();
|
||||
if (!loginOk) {
|
||||
console.log('\n⛔ 登录失败，终止测试');
|
||||
process.exit(1);
|
||||
}
|
||||
|
||||
const sessionOk = await testCreateSessionWithUpload();
|
||||
if (!sessionOk) {
|
||||
console.log('\n⛔ 会话创建失败，终止测试');
|
||||
process.exit(1);
|
||||
}
|
||||
|
||||
await testDataProfile();
|
||||
await testIntentParsing();
|
||||
await testQtoPPipeline();
|
||||
|
||||
// 汇总
|
||||
console.log(`\n${'═'.repeat(60)}`);
|
||||
console.log(`📊 测试结果汇总：${passed} 通过 / ${failed} 失败 / ${skipped} 跳过 / ${passed + failed + skipped} 总计`);
|
||||
if (failed === 0) {
|
||||
console.log('🎉 全部通过！Phase Q+P 端到端验证成功。');
|
||||
} else {
|
||||
console.log(`⚠️ 有 ${failed} 个测试失败，请检查上方输出。`);
|
||||
}
|
||||
console.log(`\n📝 测试会话 ID: ${sessionId}（可在数据库中查询详情）`);
|
||||
console.log('═'.repeat(60));
|
||||
|
||||
process.exit(failed > 0 ? 1 : 0);
|
||||
}
|
||||
|
||||
main().catch(e => {
|
||||
console.error('💥 测试脚本异常:', e);
|
||||
process.exit(1);
|
||||
});
|
||||
395
backend/scripts/test-ssa-planner-pipeline.ts
Normal file
395
backend/scripts/test-ssa-planner-pipeline.ts
Normal file
@@ -0,0 +1,395 @@
|
||||
/**
|
||||
* SSA Phase P — Tracer Bullet 测试脚本
|
||||
*
|
||||
* 验证范围：
|
||||
* 1. ConfigLoader 加载 + Zod 校验 3 个 JSON 配置文件
|
||||
* 2. DecisionTableService 四维匹配（6 种场景）
|
||||
* 3. FlowTemplateService 模板填充 + EPV 截断
|
||||
* 4. Q→P 集成：mock ParsedQuery → WorkflowPlan
|
||||
*
|
||||
* 运行方式：npx tsx scripts/test-ssa-planner-pipeline.ts
|
||||
* 不依赖：数据库、LLM、R 引擎
|
||||
*/
|
||||
|
||||
import { toolsRegistryLoader, decisionTablesLoader, flowTemplatesLoader, reloadAllConfigs } from '../src/modules/ssa/config/index.js';
|
||||
import { decisionTableService } from '../src/modules/ssa/services/DecisionTableService.js';
|
||||
import { flowTemplateService } from '../src/modules/ssa/services/FlowTemplateService.js';
|
||||
import type { ParsedQuery } from '../src/modules/ssa/types/query.types.js';
|
||||
import type { DataProfile } from '../src/modules/ssa/services/DataProfileService.js';
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 工具函数
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
let passed = 0;
|
||||
let failed = 0;
|
||||
|
||||
function assert(condition: boolean, testName: string, detail?: string) {
|
||||
if (condition) {
|
||||
console.log(` ✅ ${testName}`);
|
||||
passed++;
|
||||
} else {
|
||||
console.log(` ❌ ${testName}${detail ? ` — ${detail}` : ''}`);
|
||||
failed++;
|
||||
}
|
||||
}
|
||||
|
||||
function section(title: string) {
|
||||
console.log(`\n${'─'.repeat(60)}`);
|
||||
console.log(`📋 ${title}`);
|
||||
console.log('─'.repeat(60));
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// Mock 数据
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
function makeParsedQuery(overrides: Partial<ParsedQuery>): ParsedQuery {
|
||||
return {
|
||||
goal: 'comparison',
|
||||
outcome_var: 'BP',
|
||||
outcome_type: 'continuous',
|
||||
predictor_vars: ['Drug'],
|
||||
predictor_types: ['binary'],
|
||||
grouping_var: 'Drug',
|
||||
design: 'independent',
|
||||
confidence: 0.9,
|
||||
reasoning: 'test',
|
||||
needsClarification: false,
|
||||
...overrides,
|
||||
};
|
||||
}
|
||||
|
||||
function makeMockProfile(outcomeVar: string, minEventCount: number): DataProfile {
|
||||
return {
|
||||
totalRows: 200,
|
||||
totalColumns: 10,
|
||||
columns: [
|
||||
{
|
||||
name: outcomeVar,
|
||||
type: 'categorical',
|
||||
missing: 0,
|
||||
missingPercent: 0,
|
||||
unique: 2,
|
||||
topValues: [
|
||||
{ value: '0', count: 200 - minEventCount },
|
||||
{ value: '1', count: minEventCount },
|
||||
],
|
||||
},
|
||||
{ name: 'Age', type: 'numeric', missing: 0, missingPercent: 0, unique: 50 },
|
||||
{ name: 'Sex', type: 'categorical', missing: 0, missingPercent: 0, unique: 2 },
|
||||
{ name: 'BMI', type: 'numeric', missing: 5, missingPercent: 2.5, unique: 80 },
|
||||
{ name: 'Smoking', type: 'categorical', missing: 0, missingPercent: 0, unique: 2 },
|
||||
{ name: 'SBP', type: 'numeric', missing: 0, missingPercent: 0, unique: 100 },
|
||||
],
|
||||
} as any;
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 1: ConfigLoader + Zod 校验
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
function testConfigLoading() {
|
||||
section('测试 1: ConfigLoader 加载 + Zod 校验');
|
||||
|
||||
try {
|
||||
const tools = toolsRegistryLoader.get();
|
||||
assert(!!tools, '工具注册表加载成功');
|
||||
assert(tools.tools.length >= 7, `工具数量 >= 7（实际 ${tools.tools.length}）`);
|
||||
assert(tools.tools.every(t => /^ST_[A-Z_]+$/.test(t.code)), '所有工具 code 格式正确 (ST_XXX)');
|
||||
|
||||
const toolCodes = tools.tools.map(t => t.code);
|
||||
assert(toolCodes.includes('ST_DESCRIPTIVE'), '包含 ST_DESCRIPTIVE');
|
||||
assert(toolCodes.includes('ST_T_TEST_IND'), '包含 ST_T_TEST_IND');
|
||||
assert(toolCodes.includes('ST_LOGISTIC_BINARY'), '包含 ST_LOGISTIC_BINARY');
|
||||
} catch (e: any) {
|
||||
assert(false, '工具注册表加载失败', e.message);
|
||||
}
|
||||
|
||||
try {
|
||||
const rules = decisionTablesLoader.get();
|
||||
assert(!!rules, '决策表加载成功');
|
||||
assert(rules.length >= 9, `规则数量 >= 9（实际 ${rules.length}）`);
|
||||
assert(rules.every(r => r.id && r.goal && r.primaryTool), '所有规则含必填字段');
|
||||
|
||||
const ids = rules.map(r => r.id);
|
||||
assert(ids.includes('DESC_ANY'), '包含 DESC_ANY 兜底规则');
|
||||
assert(ids.includes('COHORT_STUDY'), '包含 COHORT_STUDY 队列研究规则');
|
||||
} catch (e: any) {
|
||||
assert(false, '决策表加载失败', e.message);
|
||||
}
|
||||
|
||||
try {
|
||||
const templates = flowTemplatesLoader.get();
|
||||
assert(!!templates, '流程模板加载成功');
|
||||
assert(templates.templates.length >= 5, `模板数量 >= 5（实际 ${templates.templates.length}）`);
|
||||
|
||||
const ids = templates.templates.map(t => t.id);
|
||||
assert(ids.includes('standard_analysis'), '包含 standard_analysis 模板');
|
||||
assert(ids.includes('cohort_study_standard'), '包含 cohort_study_standard 模板');
|
||||
assert(ids.includes('descriptive_only'), '包含 descriptive_only 模板');
|
||||
} catch (e: any) {
|
||||
assert(false, '流程模板加载失败', e.message);
|
||||
}
|
||||
|
||||
// 热更新测试
|
||||
try {
|
||||
const results = reloadAllConfigs();
|
||||
assert(results.every(r => r.success), `热更新全部成功（${results.length} 个文件）`);
|
||||
} catch (e: any) {
|
||||
assert(false, '热更新失败', e.message);
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 2: DecisionTableService 四维匹配
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
function testDecisionTableMatching() {
|
||||
section('测试 2: DecisionTableService 四维匹配');
|
||||
|
||||
// 场景 A: 两组连续变量差异比较（独立样本）→ T 检验 + Mann-Whitney fallback
|
||||
const queryA = makeParsedQuery({
|
||||
goal: 'comparison',
|
||||
outcome_type: 'continuous',
|
||||
predictor_types: ['binary'],
|
||||
design: 'independent',
|
||||
});
|
||||
const matchA = decisionTableService.match(queryA);
|
||||
assert(matchA.primaryTool === 'ST_T_TEST_IND', `场景 A: Primary = ST_T_TEST_IND（实际 ${matchA.primaryTool}）`);
|
||||
assert(matchA.fallbackTool === 'ST_MANN_WHITNEY', `场景 A: Fallback = ST_MANN_WHITNEY（实际 ${matchA.fallbackTool}）`);
|
||||
assert(matchA.switchCondition !== null, '场景 A: 有 switchCondition（正态性检验）');
|
||||
assert(matchA.templateId === 'standard_analysis', `场景 A: Template = standard_analysis（实际 ${matchA.templateId}）`);
|
||||
|
||||
// 场景 B: 配对设计
|
||||
const queryB = makeParsedQuery({
|
||||
goal: 'comparison',
|
||||
outcome_type: 'continuous',
|
||||
predictor_types: ['binary'],
|
||||
design: 'paired',
|
||||
});
|
||||
const matchB = decisionTableService.match(queryB);
|
||||
assert(matchB.primaryTool === 'ST_T_TEST_PAIRED', `场景 B: Primary = ST_T_TEST_PAIRED（实际 ${matchB.primaryTool}）`);
|
||||
assert(matchB.templateId === 'paired_analysis', `场景 B: Template = paired_analysis（实际 ${matchB.templateId}）`);
|
||||
|
||||
// 场景 C: 分类 vs 分类 → 卡方检验
|
||||
const queryC = makeParsedQuery({
|
||||
goal: 'comparison',
|
||||
outcome_type: 'categorical',
|
||||
predictor_types: ['categorical'],
|
||||
design: 'independent',
|
||||
});
|
||||
const matchC = decisionTableService.match(queryC);
|
||||
assert(matchC.primaryTool === 'ST_CHI_SQUARE', `场景 C: Primary = ST_CHI_SQUARE（实际 ${matchC.primaryTool}）`);
|
||||
|
||||
// 场景 D: 相关分析（连续 vs 连续）
|
||||
const queryD = makeParsedQuery({
|
||||
goal: 'correlation',
|
||||
outcome_type: 'continuous',
|
||||
predictor_types: ['continuous'],
|
||||
design: 'independent',
|
||||
});
|
||||
const matchD = decisionTableService.match(queryD);
|
||||
assert(matchD.primaryTool === 'ST_CORRELATION', `场景 D: Primary = ST_CORRELATION（实际 ${matchD.primaryTool}）`);
|
||||
|
||||
// 场景 E: Logistic 回归
|
||||
const queryE = makeParsedQuery({
|
||||
goal: 'regression',
|
||||
outcome_type: 'binary',
|
||||
predictor_types: ['continuous'],
|
||||
design: 'independent',
|
||||
});
|
||||
const matchE = decisionTableService.match(queryE);
|
||||
assert(matchE.primaryTool === 'ST_LOGISTIC_BINARY', `场景 E: Primary = ST_LOGISTIC_BINARY（实际 ${matchE.primaryTool}）`);
|
||||
|
||||
// 场景 F: 描述统计 fallback
|
||||
const queryF = makeParsedQuery({
|
||||
goal: 'descriptive',
|
||||
outcome_type: null,
|
||||
predictor_types: [],
|
||||
});
|
||||
const matchF = decisionTableService.match(queryF);
|
||||
assert(matchF.primaryTool === 'ST_DESCRIPTIVE', `场景 F: Primary = ST_DESCRIPTIVE（实际 ${matchF.primaryTool}）`);
|
||||
|
||||
// 场景 G: 队列研究
|
||||
const queryG = makeParsedQuery({
|
||||
goal: 'cohort_study',
|
||||
outcome_type: 'binary',
|
||||
predictor_types: ['categorical'],
|
||||
design: 'independent',
|
||||
});
|
||||
const matchG = decisionTableService.match(queryG);
|
||||
assert(matchG.templateId === 'cohort_study_standard', `场景 G: Template = cohort_study_standard（实际 ${matchG.templateId}）`);
|
||||
|
||||
// 场景 H: 未知 goal → 应该 fallback 到描述统计
|
||||
const queryH = makeParsedQuery({
|
||||
goal: 'descriptive' as any, // 模拟未匹配场景
|
||||
outcome_type: 'datetime' as any,
|
||||
predictor_types: ['datetime' as any],
|
||||
});
|
||||
const matchH = decisionTableService.match(queryH);
|
||||
assert(matchH.primaryTool === 'ST_DESCRIPTIVE', `场景 H: 无精确匹配 → Fallback ST_DESCRIPTIVE（实际 ${matchH.primaryTool}）`);
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 3: FlowTemplateService 模板填充
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
function testFlowTemplateFilling() {
|
||||
section('测试 3: FlowTemplateService 模板填充');
|
||||
|
||||
// 场景 A: standard_analysis（有 fallback → 3 步）
|
||||
const queryA = makeParsedQuery({
|
||||
goal: 'comparison',
|
||||
outcome_type: 'continuous',
|
||||
predictor_types: ['binary'],
|
||||
outcome_var: 'BP',
|
||||
predictor_vars: ['Drug'],
|
||||
grouping_var: 'Drug',
|
||||
});
|
||||
const matchA = decisionTableService.match(queryA);
|
||||
const fillA = flowTemplateService.fill(matchA, queryA);
|
||||
assert(fillA.steps.length === 3, `场景 A: 3 步流程（实际 ${fillA.steps.length}）`);
|
||||
assert(fillA.steps[0].toolCode === 'ST_DESCRIPTIVE', `场景 A 步骤 1: ST_DESCRIPTIVE（实际 ${fillA.steps[0].toolCode}）`);
|
||||
assert(fillA.steps[1].toolCode === 'ST_T_TEST_IND', `场景 A 步骤 2: ST_T_TEST_IND（实际 ${fillA.steps[1].toolCode}）`);
|
||||
assert(fillA.steps[2].toolCode === 'ST_MANN_WHITNEY', `场景 A 步骤 3: ST_MANN_WHITNEY（实际 ${fillA.steps[2].toolCode}）`);
|
||||
assert(fillA.steps[2].isSensitivity === true, '场景 A 步骤 3: isSensitivity = true');
|
||||
assert(fillA.epvWarning === null, '场景 A: 无 EPV 警告');
|
||||
|
||||
// 场景 B: descriptive_only（无 fallback → 1 步）
|
||||
const queryB = makeParsedQuery({
|
||||
goal: 'descriptive',
|
||||
outcome_type: null,
|
||||
predictor_types: [],
|
||||
});
|
||||
const matchB = decisionTableService.match(queryB);
|
||||
const fillB = flowTemplateService.fill(matchB, queryB);
|
||||
assert(fillB.steps.length === 1, `场景 B: 1 步流程（实际 ${fillB.steps.length}）`);
|
||||
assert(fillB.steps[0].toolCode === 'ST_DESCRIPTIVE', '场景 B: ST_DESCRIPTIVE');
|
||||
|
||||
// 场景 C: 队列研究 → 3 步 (Table 1/2/3)
|
||||
const queryC = makeParsedQuery({
|
||||
goal: 'cohort_study',
|
||||
outcome_var: 'Event',
|
||||
outcome_type: 'binary',
|
||||
predictor_vars: ['Age', 'Sex', 'BMI', 'Smoking', 'SBP'],
|
||||
predictor_types: ['continuous', 'binary', 'continuous', 'binary', 'continuous'],
|
||||
grouping_var: 'Drug',
|
||||
design: 'independent',
|
||||
});
|
||||
const matchC = decisionTableService.match(queryC);
|
||||
const fillC = flowTemplateService.fill(matchC, queryC);
|
||||
assert(fillC.steps.length === 3, `场景 C: 队列研究 3 步（实际 ${fillC.steps.length}）`);
|
||||
assert(fillC.steps.length > 0 && fillC.steps[0].name.includes('表1'), `场景 C 步骤 1: 表1（实际 "${fillC.steps[0]?.name ?? 'N/A'}"）`);
|
||||
assert(fillC.steps.length > 1 && fillC.steps[1].name.includes('表2'), `场景 C 步骤 2: 表2（实际 "${fillC.steps[1]?.name ?? 'N/A'}"）`);
|
||||
assert(fillC.steps.length > 2 && fillC.steps[2].name.includes('表3'), `场景 C 步骤 3: 表3（实际 "${fillC.steps[2]?.name ?? 'N/A'}"）`);
|
||||
|
||||
// 场景 D: EPV 截断 — 30 个事件 / 10 = 最多 3 个变量
|
||||
const queryD = makeParsedQuery({
|
||||
goal: 'cohort_study',
|
||||
outcome_var: 'Event',
|
||||
outcome_type: 'binary',
|
||||
predictor_vars: ['Age', 'Sex', 'BMI', 'Smoking', 'SBP', 'HR', 'Chol', 'LDL'],
|
||||
predictor_types: ['continuous', 'binary', 'continuous', 'binary', 'continuous', 'continuous', 'continuous', 'continuous'],
|
||||
grouping_var: 'Drug',
|
||||
design: 'independent',
|
||||
});
|
||||
const profileD = makeMockProfile('Event', 30); // 只有 30 个 event → max 3 vars
|
||||
const matchD = decisionTableService.match(queryD);
|
||||
const fillD = flowTemplateService.fill(matchD, queryD, profileD);
|
||||
|
||||
const table3Step = fillD.steps.find(s => s.name.includes('表3'));
|
||||
if (table3Step) {
|
||||
const predictors = table3Step.params.predictors as string[] | undefined;
|
||||
if (predictors) {
|
||||
assert(predictors.length <= 3, `场景 D EPV 截断: 自变量 <= 3（实际 ${predictors.length}，原始 8）`);
|
||||
} else {
|
||||
assert(false, '场景 D EPV 截断: 未找到 predictors 参数');
|
||||
}
|
||||
} else {
|
||||
assert(false, '场景 D: 未找到表3 步骤');
|
||||
}
|
||||
assert(fillD.epvWarning !== null, `场景 D: 有 EPV 警告（${fillD.epvWarning?.substring(0, 40)}...）`);
|
||||
|
||||
// 场景 E: 配对分析 → 2 步（无 sensitivity）
|
||||
const queryE = makeParsedQuery({
|
||||
goal: 'comparison',
|
||||
outcome_type: 'continuous',
|
||||
predictor_types: ['binary'],
|
||||
design: 'paired',
|
||||
outcome_var: 'BP_after',
|
||||
predictor_vars: ['BP_before'],
|
||||
});
|
||||
const matchE = decisionTableService.match(queryE);
|
||||
const fillE = flowTemplateService.fill(matchE, queryE);
|
||||
assert(fillE.steps.length === 2, `场景 E: 配对分析 2 步（实际 ${fillE.steps.length}）`);
|
||||
assert(fillE.steps.every(s => !s.isSensitivity), '场景 E: 无敏感性分析步骤');
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 4: PlannedTrace 完整性
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
function testPlannedTrace() {
|
||||
section('测试 4: PlannedTrace 数据完整性');
|
||||
|
||||
const query = makeParsedQuery({
|
||||
goal: 'comparison',
|
||||
outcome_type: 'continuous',
|
||||
predictor_types: ['binary'],
|
||||
design: 'independent',
|
||||
outcome_var: 'BP',
|
||||
predictor_vars: ['Drug'],
|
||||
grouping_var: 'Drug',
|
||||
});
|
||||
|
||||
const match = decisionTableService.match(query);
|
||||
const fill = flowTemplateService.fill(match, query);
|
||||
|
||||
// PlannedTrace 应具备的信息
|
||||
assert(match.rule.id !== '', 'PlannedTrace: matchedRule 非空');
|
||||
assert(match.primaryTool === 'ST_T_TEST_IND', `PlannedTrace: primaryTool = ST_T_TEST_IND`);
|
||||
assert(match.fallbackTool === 'ST_MANN_WHITNEY', `PlannedTrace: fallbackTool = ST_MANN_WHITNEY`);
|
||||
assert(match.switchCondition !== null, 'PlannedTrace: switchCondition 非空');
|
||||
assert(fill.templateId === 'standard_analysis', 'PlannedTrace: templateUsed = standard_analysis');
|
||||
assert(match.matchScore > 0, `PlannedTrace: matchScore > 0（实际 ${match.matchScore}）`);
|
||||
|
||||
// 确认参数正确传递
|
||||
const primaryStep = fill.steps.find(s => s.role === 'primary_test');
|
||||
assert(!!primaryStep, 'Primary step 存在');
|
||||
if (primaryStep) {
|
||||
assert(primaryStep.params.group_var === 'Drug' || primaryStep.params.value_var === 'BP',
|
||||
`Primary step 参数包含正确变量`);
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 运行所有测试
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
console.log('\n🧪 SSA Phase P — Tracer Bullet 测试\n');
|
||||
console.log('测试范围：ConfigLoader → DecisionTable → FlowTemplate → PlannedTrace');
|
||||
console.log('依赖项：无（不需要数据库、LLM、R 引擎）\n');
|
||||
|
||||
try {
|
||||
testConfigLoading();
|
||||
testDecisionTableMatching();
|
||||
testFlowTemplateFilling();
|
||||
testPlannedTrace();
|
||||
} catch (e: any) {
|
||||
console.error(`\n💥 测试过程中发生未捕获异常：${e.message}`);
|
||||
console.error(e.stack);
|
||||
failed++;
|
||||
}
|
||||
|
||||
// 汇总
|
||||
console.log(`\n${'═'.repeat(60)}`);
|
||||
console.log(`📊 测试结果汇总：${passed} 通过 / ${failed} 失败 / ${passed + failed} 总计`);
|
||||
if (failed === 0) {
|
||||
console.log('🎉 全部通过！P 层 Pipeline 验证成功。');
|
||||
} else {
|
||||
console.log(`⚠️ 有 ${failed} 个测试失败，请检查上方输出。`);
|
||||
}
|
||||
console.log('═'.repeat(60));
|
||||
|
||||
process.exit(failed > 0 ? 1 : 0);
|
||||
663
backend/scripts/test-ssa-qper-e2e.ts
Normal file
663
backend/scripts/test-ssa-qper-e2e.ts
Normal file
@@ -0,0 +1,663 @@
|
||||
/**
|
||||
* SSA Q→P→E→R — 完整 QPER 链路端到端集成测试
|
||||
*
|
||||
* 测试链路：
|
||||
* 登录 → 创建会话+上传 CSV → 数据画像
|
||||
* → Q 层（LLM Intent）→ P 层（Plan）
|
||||
* → E 层（R 引擎执行）→ R 层（LLM 结论生成）
|
||||
* → 结论 API 缓存验证
|
||||
*
|
||||
* 依赖：Node.js 后端 + PostgreSQL + Python extraction_service + R 引擎 + LLM 服务
|
||||
* 运行方式：npx tsx scripts/test-ssa-qper-e2e.ts
|
||||
*
|
||||
* 测试数据：docs/03-业务模块/SSA-智能统计分析/05-测试文档/test.csv
|
||||
* 测试用户：13800000001 / 123456
|
||||
*/
|
||||
|
||||
import { readFileSync } from 'fs';
|
||||
import { join, dirname } from 'path';
|
||||
import { fileURLToPath } from 'url';
|
||||
|
||||
const __filename = fileURLToPath(import.meta.url);
|
||||
const __dirname = dirname(__filename);
|
||||
|
||||
const BASE_URL = 'http://localhost:3000';
|
||||
const TEST_PHONE = '13800000001';
|
||||
const TEST_PASSWORD = '123456';
|
||||
const TEST_CSV_PATH = join(__dirname, '../../docs/03-业务模块/SSA-智能统计分析/05-测试文档/test.csv');
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 工具函数
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
let passed = 0;
|
||||
let failed = 0;
|
||||
let skipped = 0;
|
||||
let token = '';
|
||||
let sessionId = '';
|
||||
|
||||
function assert(condition: boolean, testName: string, detail?: string) {
|
||||
if (condition) {
|
||||
console.log(` ✅ ${testName}`);
|
||||
passed++;
|
||||
} else {
|
||||
console.log(` ❌ ${testName}${detail ? ` — ${detail}` : ''}`);
|
||||
failed++;
|
||||
}
|
||||
}
|
||||
|
||||
function skip(testName: string, reason: string) {
|
||||
console.log(` ⏭️ ${testName} — 跳过：${reason}`);
|
||||
skipped++;
|
||||
}
|
||||
|
||||
function section(title: string) {
|
||||
console.log(`\n${'─'.repeat(60)}`);
|
||||
console.log(`📋 ${title}`);
|
||||
console.log('─'.repeat(60));
|
||||
}
|
||||
|
||||
function authHeaders(contentType?: string): Record<string, string> {
|
||||
const headers: Record<string, string> = {
|
||||
'Authorization': `Bearer ${token}`,
|
||||
};
|
||||
if (contentType) {
|
||||
headers['Content-Type'] = contentType;
|
||||
}
|
||||
return headers;
|
||||
}
|
||||
|
||||
async function apiPost(path: string, body: any, headers?: Record<string, string>): Promise<any> {
|
||||
const res = await fetch(`${BASE_URL}${path}`, {
|
||||
method: 'POST',
|
||||
headers: headers || authHeaders('application/json'),
|
||||
body: typeof body === 'string' ? body : JSON.stringify(body),
|
||||
});
|
||||
const text = await res.text();
|
||||
try {
|
||||
return { status: res.status, data: JSON.parse(text) };
|
||||
} catch {
|
||||
return { status: res.status, data: text };
|
||||
}
|
||||
}
|
||||
|
||||
async function apiGet(path: string): Promise<any> {
|
||||
const res = await fetch(`${BASE_URL}${path}`, {
|
||||
method: 'GET',
|
||||
headers: authHeaders(),
|
||||
});
|
||||
const text = await res.text();
|
||||
try {
|
||||
return { status: res.status, data: JSON.parse(text) };
|
||||
} catch {
|
||||
return { status: res.status, data: text };
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 1: 登录获取 Token
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testLogin(): Promise<boolean> {
|
||||
section('测试 1: 登录认证');
|
||||
|
||||
try {
|
||||
const res = await apiPost('/api/v1/auth/login/password', {
|
||||
phone: TEST_PHONE,
|
||||
password: TEST_PASSWORD,
|
||||
}, { 'Content-Type': 'application/json' });
|
||||
|
||||
assert(res.status === 200, `登录返回 200（实际 ${res.status}）`);
|
||||
|
||||
if (res.status === 200 && res.data) {
|
||||
token = res.data?.data?.tokens?.accessToken || res.data?.accessToken || res.data?.token || '';
|
||||
assert(token.length > 0, '获取到 JWT Token', `token 长度: ${token.length}`);
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, '登录请求失败', e.message);
|
||||
}
|
||||
|
||||
if (!token) {
|
||||
console.log('\n ⚠️ Token 获取失败，后续测试无法继续');
|
||||
return false;
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 2: 创建会话 + 上传 test.csv
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testCreateSession(): Promise<boolean> {
|
||||
section('测试 2: 创建会话 + 上传 test.csv');
|
||||
|
||||
try {
|
||||
const csvBuffer = readFileSync(TEST_CSV_PATH);
|
||||
assert(csvBuffer.length > 0, `test.csv 读取成功（${csvBuffer.length} bytes）`);
|
||||
|
||||
const formData = new FormData();
|
||||
const blob = new Blob([csvBuffer], { type: 'text/csv' });
|
||||
formData.append('file', blob, 'test.csv');
|
||||
|
||||
const res = await fetch(`${BASE_URL}/api/v1/ssa/sessions/`, {
|
||||
method: 'POST',
|
||||
headers: { 'Authorization': `Bearer ${token}` },
|
||||
body: formData,
|
||||
});
|
||||
|
||||
const data = await res.json();
|
||||
assert(res.status === 200, `创建会话返回 200（实际 ${res.status}）`);
|
||||
|
||||
if (data.sessionId) {
|
||||
sessionId = data.sessionId;
|
||||
assert(true, `会话 ID: ${sessionId}`);
|
||||
} else {
|
||||
assert(false, '未返回 sessionId');
|
||||
}
|
||||
|
||||
if (data.schema) {
|
||||
assert(data.schema.columns?.length > 0, `Schema 解析成功（${data.schema.columns?.length} 列, ${data.schema.rowCount} 行）`);
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, '创建会话失败', e.message);
|
||||
}
|
||||
|
||||
return !!sessionId;
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 3: 数据画像（Python DataProfiler）
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testDataProfile() {
|
||||
section('测试 3: 数据画像（Python DataProfiler）');
|
||||
|
||||
try {
|
||||
const res = await apiPost('/api/v1/ssa/workflow/profile', { sessionId });
|
||||
assert(res.status === 200, `画像请求返回 200（实际 ${res.status}）`);
|
||||
|
||||
if (res.data?.success) {
|
||||
const profile = res.data.profile;
|
||||
assert(!!profile, '画像数据非空');
|
||||
if (profile) {
|
||||
const rows = profile.row_count || profile.totalRows || 0;
|
||||
const cols = profile.column_count || profile.totalColumns || 0;
|
||||
assert(rows > 0, `行数: ${rows}`);
|
||||
assert(cols > 0, `列数: ${cols}`);
|
||||
}
|
||||
} else {
|
||||
assert(false, '画像生成失败', res.data?.error);
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, '画像请求异常', e.message);
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 4: Q 层 — LLM 意图解析
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testQLayer(): Promise<string | null> {
|
||||
section('测试 4: Q 层 — LLM 意图理解');
|
||||
|
||||
const query = '比较 sex 不同组的 Yqol 有没有差别';
|
||||
console.log(` Query: "${query}"`);
|
||||
|
||||
try {
|
||||
const start = Date.now();
|
||||
const res = await apiPost('/api/v1/ssa/workflow/intent', {
|
||||
sessionId,
|
||||
userQuery: query,
|
||||
});
|
||||
const elapsed = Date.now() - start;
|
||||
|
||||
assert(res.status === 200, `返回 200（实际 ${res.status}）`);
|
||||
|
||||
if (res.data?.success && res.data.intent) {
|
||||
const intent = res.data.intent;
|
||||
console.log(` 耗时: ${elapsed}ms`);
|
||||
console.log(` Goal: ${intent.goal}, Confidence: ${intent.confidence}`);
|
||||
console.log(` Y: ${intent.outcome_var}, X: ${JSON.stringify(intent.predictor_vars)}`);
|
||||
console.log(` Design: ${intent.design}, needsClarification: ${intent.needsClarification}`);
|
||||
|
||||
assert(intent.goal === 'comparison', `Goal = comparison（实际 ${intent.goal}）`);
|
||||
assert(intent.confidence >= 0.7, `高置信度 >= 0.7（实际 ${intent.confidence}）`);
|
||||
assert(!intent.needsClarification, '无需追问');
|
||||
|
||||
return intent.goal;
|
||||
} else {
|
||||
assert(false, 'Intent 解析失败', res.data?.error);
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, 'Q 层请求异常', e.message);
|
||||
}
|
||||
|
||||
return null;
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 5: P 层 — 工作流规划
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
let workflowId = '';
|
||||
|
||||
async function testPLayer(): Promise<boolean> {
|
||||
section('测试 5: P 层 — 工作流规划');
|
||||
|
||||
const query = '比较 sex 不同组的 Yqol 有没有差别';
|
||||
console.log(` Query: "${query}"`);
|
||||
|
||||
try {
|
||||
const start = Date.now();
|
||||
const res = await apiPost('/api/v1/ssa/workflow/plan', {
|
||||
sessionId,
|
||||
userQuery: query,
|
||||
});
|
||||
const elapsed = Date.now() - start;
|
||||
|
||||
assert(res.status === 200, `返回 200（实际 ${res.status}）`);
|
||||
|
||||
if (res.data?.success && res.data.plan) {
|
||||
const plan = res.data.plan;
|
||||
console.log(` 耗时: ${elapsed}ms`);
|
||||
console.log(` 标题: ${plan.title}`);
|
||||
console.log(` 步骤数: ${plan.total_steps}`);
|
||||
|
||||
workflowId = plan.workflow_id;
|
||||
assert(!!workflowId, `Workflow ID: ${workflowId}`);
|
||||
assert(plan.total_steps >= 2, `步骤数 >= 2（实际 ${plan.total_steps}）`);
|
||||
|
||||
plan.steps?.forEach((step: any, i: number) => {
|
||||
const sensitivity = step.is_sensitivity ? ' [敏感性]' : '';
|
||||
const guardrail = step.switch_condition ? ` | 护栏:${step.switch_condition}` : '';
|
||||
console.log(` 步骤 ${i + 1}: ${step.tool_name} (${step.tool_code})${sensitivity}${guardrail}`);
|
||||
});
|
||||
|
||||
if (plan.planned_trace) {
|
||||
console.log(` PlannedTrace: Primary=${plan.planned_trace.primaryTool}, Fallback=${plan.planned_trace.fallbackTool || 'null'}`);
|
||||
assert(!!plan.planned_trace.primaryTool, 'PlannedTrace 包含 primaryTool');
|
||||
}
|
||||
|
||||
return true;
|
||||
} else {
|
||||
assert(false, '规划失败', res.data?.error);
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, 'P 层请求异常', e.message);
|
||||
}
|
||||
|
||||
return false;
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 6: E 层 — R 引擎执行
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testELayer(): Promise<boolean> {
|
||||
section('测试 6: E 层 — R 引擎执行（含 R 层结论生成）');
|
||||
|
||||
if (!workflowId) {
|
||||
skip('E 层执行', '无 workflowId');
|
||||
return false;
|
||||
}
|
||||
|
||||
console.log(` Workflow ID: ${workflowId}`);
|
||||
console.log(` Session ID: ${sessionId}`);
|
||||
|
||||
try {
|
||||
const start = Date.now();
|
||||
const res = await apiPost(`/api/v1/ssa/workflow/${workflowId}/execute`, {
|
||||
sessionId,
|
||||
});
|
||||
const elapsed = Date.now() - start;
|
||||
|
||||
assert(res.status === 200, `返回 200（实际 ${res.status}）`);
|
||||
|
||||
if (res.data?.success && res.data.result) {
|
||||
const result = res.data.result;
|
||||
console.log(` 耗时: ${elapsed}ms`);
|
||||
console.log(` 状态: ${result.status}`);
|
||||
console.log(` 总步骤: ${result.totalSteps}, 成功: ${result.successSteps}, 完成: ${result.completedSteps}`);
|
||||
|
||||
assert(
|
||||
result.status === 'completed' || result.status === 'partial',
|
||||
`执行状态正常（${result.status}）`,
|
||||
result.status === 'error' ? '全部步骤失败' : undefined,
|
||||
);
|
||||
|
||||
assert(result.successSteps > 0, `至少 1 个步骤成功（实际 ${result.successSteps}）`);
|
||||
|
||||
// 逐步骤检查
|
||||
if (result.results && Array.isArray(result.results)) {
|
||||
for (const step of result.results) {
|
||||
const icon = step.status === 'success' || step.status === 'warning' ? '✅' : '❌';
|
||||
const pVal = step.result?.p_value != null ? `, P=${step.result.p_value_fmt || step.result.p_value}` : '';
|
||||
const blocks = step.reportBlocks?.length || 0;
|
||||
const errMsg = step.error ? ` | 错误: ${step.error.userHint || step.error.message}` : '';
|
||||
console.log(` ${icon} 步骤 ${step.stepOrder}: ${step.toolName} [${step.status}] (${step.executionMs}ms${pVal}, ${blocks} blocks${errMsg})`);
|
||||
}
|
||||
}
|
||||
|
||||
// 检查 report_blocks
|
||||
if (result.reportBlocks && result.reportBlocks.length > 0) {
|
||||
assert(true, `聚合 reportBlocks: ${result.reportBlocks.length} 个`);
|
||||
const types = result.reportBlocks.map((b: any) => b.type);
|
||||
const uniqueTypes = [...new Set(types)];
|
||||
console.log(` Block 类型分布: ${uniqueTypes.join(', ')}`);
|
||||
}
|
||||
|
||||
// 检查 R 层结论
|
||||
if (result.conclusion) {
|
||||
console.log('\n ── R 层结论验证 ──');
|
||||
const c = result.conclusion;
|
||||
|
||||
assert(!!c.executive_summary, `executive_summary 非空（${c.executive_summary?.length || 0} 字）`);
|
||||
assert(Array.isArray(c.key_findings) && c.key_findings.length > 0,
|
||||
`key_findings 非空（${c.key_findings?.length || 0} 条）`);
|
||||
assert(!!c.statistical_summary, 'statistical_summary 存在');
|
||||
assert(Array.isArray(c.limitations) && c.limitations.length > 0,
|
||||
`limitations 非空（${c.limitations?.length || 0} 条）`);
|
||||
assert(!!c.generated_at, `generated_at: ${c.generated_at}`);
|
||||
assert(!!c.source, `source: ${c.source}`);
|
||||
|
||||
// 打印结论内容摘要
|
||||
console.log(` 结论来源: ${c.source === 'llm' ? 'AI 智能生成' : '规则引擎'}`);
|
||||
console.log(` 摘要前 200 字: ${c.executive_summary?.substring(0, 200)}...`);
|
||||
|
||||
if (c.key_findings?.length > 0) {
|
||||
console.log(' 主要发现:');
|
||||
c.key_findings.slice(0, 3).forEach((f: string, i: number) => {
|
||||
console.log(` ${i + 1}. ${f.substring(0, 120)}`);
|
||||
});
|
||||
}
|
||||
|
||||
if (c.statistical_summary) {
|
||||
console.log(` 统计概览: ${c.statistical_summary.total_tests} 项检验, ${c.statistical_summary.significant_results} 项显著`);
|
||||
console.log(` 使用方法: ${c.statistical_summary.methods_used?.join(', ')}`);
|
||||
}
|
||||
|
||||
if (c.step_summaries?.length > 0) {
|
||||
console.log(' 步骤摘要:');
|
||||
c.step_summaries.forEach((s: any) => {
|
||||
const sig = s.is_significant ? ' (显著*)' : '';
|
||||
console.log(` 步骤${s.step_number} ${s.tool_name}: ${s.summary?.substring(0, 100)}${sig}`);
|
||||
});
|
||||
}
|
||||
|
||||
if (c.limitations?.length > 0) {
|
||||
console.log(' 局限性:');
|
||||
c.limitations.slice(0, 3).forEach((l: string, i: number) => {
|
||||
console.log(` ${i + 1}. ${l.substring(0, 120)}`);
|
||||
});
|
||||
}
|
||||
|
||||
if (c.recommendations?.length > 0) {
|
||||
console.log(' 建议:');
|
||||
c.recommendations.slice(0, 2).forEach((r: string, i: number) => {
|
||||
console.log(` ${i + 1}. ${r.substring(0, 120)}`);
|
||||
});
|
||||
}
|
||||
|
||||
// 验证 workflow_id 一致
|
||||
if (c.workflow_id) {
|
||||
assert(c.workflow_id === workflowId, `conclusion.workflow_id 与 workflowId 一致`);
|
||||
}
|
||||
} else {
|
||||
assert(false, 'R 层未返回 conclusion');
|
||||
}
|
||||
|
||||
return result.successSteps > 0;
|
||||
} else {
|
||||
assert(false, '执行失败', res.data?.error);
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, 'E 层请求异常', e.message);
|
||||
}
|
||||
|
||||
return false;
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 7: 结论 API 缓存验证
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testConclusionAPI() {
|
||||
section('测试 7: 结论 API + 缓存验证');
|
||||
|
||||
if (!sessionId) {
|
||||
skip('结论 API', '无 sessionId');
|
||||
return;
|
||||
}
|
||||
|
||||
try {
|
||||
const start = Date.now();
|
||||
const res = await apiGet(`/api/v1/ssa/workflow/sessions/${sessionId}/conclusion`);
|
||||
const elapsed = Date.now() - start;
|
||||
|
||||
assert(res.status === 200, `返回 200（实际 ${res.status}）`);
|
||||
|
||||
if (res.data?.success && res.data.conclusion) {
|
||||
const c = res.data.conclusion;
|
||||
console.log(` 耗时: ${elapsed}ms`);
|
||||
console.log(` 来源: ${res.data.source}`);
|
||||
|
||||
assert(!!c.executive_summary, 'executive_summary 非空');
|
||||
assert(Array.isArray(c.key_findings), 'key_findings 是数组');
|
||||
assert(!!c.generated_at, `generated_at: ${c.generated_at}`);
|
||||
|
||||
// 二次调用验证缓存
|
||||
console.log('\n ── 缓存验证（二次调用） ──');
|
||||
const start2 = Date.now();
|
||||
const res2 = await apiGet(`/api/v1/ssa/workflow/sessions/${sessionId}/conclusion`);
|
||||
const elapsed2 = Date.now() - start2;
|
||||
|
||||
assert(res2.status === 200, '二次调用返回 200');
|
||||
console.log(` 二次调用耗时: ${elapsed2}ms`);
|
||||
|
||||
if (elapsed2 < elapsed && res.data.source === 'cache') {
|
||||
assert(true, `缓存命中（${elapsed2}ms << ${elapsed}ms）`);
|
||||
} else {
|
||||
console.log(` ℹ️ 首次 ${elapsed}ms, 二次 ${elapsed2}ms（缓存效果取决于实现）`);
|
||||
}
|
||||
} else if (res.status === 404) {
|
||||
skip('结论 API', '未找到已完成的 workflow（可能是 E 层全部失败）');
|
||||
} else {
|
||||
assert(false, '获取结论失败', res.data?.error || JSON.stringify(res.data).substring(0, 200));
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, '结论 API 异常', e.message);
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 8: 第二条链路（相关分析 Q→P→E→R）
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testSecondScenario() {
|
||||
section('测试 8: 第二条完整链路（相关分析 age vs bmi）');
|
||||
|
||||
const query = '分析 age 和 bmi 的相关性';
|
||||
console.log(` Query: "${query}"`);
|
||||
|
||||
try {
|
||||
// Q → P: Plan
|
||||
const planRes = await apiPost('/api/v1/ssa/workflow/plan', {
|
||||
sessionId,
|
||||
userQuery: query,
|
||||
});
|
||||
|
||||
assert(planRes.status === 200, 'Plan 返回 200');
|
||||
|
||||
if (!planRes.data?.success || !planRes.data.plan) {
|
||||
assert(false, 'Plan 失败', planRes.data?.error);
|
||||
return;
|
||||
}
|
||||
|
||||
const plan = planRes.data.plan;
|
||||
const wfId = plan.workflow_id;
|
||||
console.log(` Workflow: ${wfId}, 步骤数: ${plan.total_steps}`);
|
||||
plan.steps?.forEach((s: any, i: number) => {
|
||||
console.log(` 步骤 ${i + 1}: ${s.tool_name} (${s.tool_code})`);
|
||||
});
|
||||
|
||||
// P → E → R: Execute
|
||||
const start = Date.now();
|
||||
const execRes = await apiPost(`/api/v1/ssa/workflow/${wfId}/execute`, { sessionId });
|
||||
const elapsed = Date.now() - start;
|
||||
|
||||
assert(execRes.status === 200, 'Execute 返回 200');
|
||||
|
||||
if (execRes.data?.success && execRes.data.result) {
|
||||
const result = execRes.data.result;
|
||||
console.log(` 执行耗时: ${elapsed}ms, 状态: ${result.status}, 成功步骤: ${result.successSteps}/${result.totalSteps}`);
|
||||
|
||||
assert(result.successSteps > 0, `至少 1 步成功（实际 ${result.successSteps}）`);
|
||||
|
||||
for (const step of (result.results || [])) {
|
||||
const icon = step.status === 'success' || step.status === 'warning' ? '✅' : '❌';
|
||||
const pVal = step.result?.p_value != null ? `, P=${step.result.p_value_fmt || step.result.p_value}` : '';
|
||||
console.log(` ${icon} 步骤 ${step.stepOrder}: ${step.toolName} [${step.status}] (${step.executionMs}ms${pVal})`);
|
||||
}
|
||||
|
||||
// 验证 R 层结论
|
||||
if (result.conclusion) {
|
||||
const c = result.conclusion;
|
||||
assert(!!c.executive_summary, `R 层结论存在（来源: ${c.source}）`);
|
||||
console.log(` 结论摘要: ${c.executive_summary?.substring(0, 150)}...`);
|
||||
|
||||
// 相关分析应该提到相关系数
|
||||
const mentionsCorrelation =
|
||||
c.executive_summary?.includes('相关') ||
|
||||
c.executive_summary?.includes('correlation') ||
|
||||
c.executive_summary?.includes('r =') ||
|
||||
c.executive_summary?.includes('r=');
|
||||
if (mentionsCorrelation) {
|
||||
assert(true, '结论中提到了相关性分析');
|
||||
} else {
|
||||
console.log(' ℹ️ 结论未明确提到"相关"（可能是 fallback 结论）');
|
||||
}
|
||||
} else {
|
||||
skip('R 层结论', '未返回 conclusion');
|
||||
}
|
||||
} else {
|
||||
assert(false, '执行失败', execRes.data?.error);
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, '第二条链路异常', e.message);
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 测试 9: 错误分类验证（E_COLUMN_NOT_FOUND 等）
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function testErrorClassification() {
|
||||
section('测试 9: E 层错误分类验证（构造异常查询）');
|
||||
|
||||
const query = '比较 NONEXISTENT_GROUP 不同组的 FAKE_OUTCOME';
|
||||
console.log(` 构造异常 Query: "${query}"`);
|
||||
console.log(' ℹ️ 此测试验证 LLM 面对不存在的变量名时的行为');
|
||||
|
||||
try {
|
||||
const res = await apiPost('/api/v1/ssa/workflow/intent', {
|
||||
sessionId,
|
||||
userQuery: query,
|
||||
});
|
||||
|
||||
if (res.data?.success && res.data.intent) {
|
||||
const intent = res.data.intent;
|
||||
console.log(` LLM 返回: goal=${intent.goal}, confidence=${intent.confidence}`);
|
||||
console.log(` Y=${intent.outcome_var}, X=${JSON.stringify(intent.predictor_vars)}`);
|
||||
|
||||
// Zod 动态校验应该拦截不存在的变量名
|
||||
// 或者 LLM 会给出低置信度
|
||||
if (intent.confidence < 0.7 || intent.needsClarification) {
|
||||
assert(true, `LLM 识别到异常（confidence=${intent.confidence}）或触发追问`);
|
||||
} else {
|
||||
console.log(' ℹ️ LLM 未识别到异常变量，可能猜测了现有变量作为替代');
|
||||
}
|
||||
} else {
|
||||
// Intent 解析失败也是可以接受的（Zod 拦截了幻觉变量）
|
||||
console.log(` Intent 解析结果: ${res.data?.error || '失败/降级'}`);
|
||||
assert(true, '异常输入被处理（未崩溃）');
|
||||
}
|
||||
} catch (e: any) {
|
||||
assert(false, '异常查询处理失败（不应崩溃）', e.message);
|
||||
}
|
||||
}
|
||||
|
||||
// ────────────────────────────────────────────
|
||||
// 运行所有测试
|
||||
// ────────────────────────────────────────────
|
||||
|
||||
async function main() {
|
||||
console.log('\n🧪 SSA QPER — 完整链路端到端集成测试（Q→P→E→R）\n');
|
||||
console.log('测试链路：登录 → 上传 CSV → 画像 → Q(Intent) → P(Plan) → E(Execute) → R(Conclusion)');
|
||||
console.log(`测试用户：${TEST_PHONE}`);
|
||||
console.log(`后端地址：${BASE_URL}`);
|
||||
console.log(`测试文件：${TEST_CSV_PATH}\n`);
|
||||
|
||||
// 前置检查
|
||||
try {
|
||||
readFileSync(TEST_CSV_PATH);
|
||||
} catch {
|
||||
console.error('❌ test.csv 文件不存在，请检查路径');
|
||||
process.exit(1);
|
||||
}
|
||||
|
||||
try {
|
||||
const health = await fetch(`${BASE_URL}/health`).catch(() => null);
|
||||
if (!health || health.status !== 200) {
|
||||
console.error('❌ 后端服务未启动');
|
||||
process.exit(1);
|
||||
}
|
||||
console.log('✅ 后端服务可达');
|
||||
} catch {
|
||||
console.error('❌ 后端服务不可达');
|
||||
process.exit(1);
|
||||
}
|
||||
|
||||
// 顺序执行
|
||||
const loginOk = await testLogin();
|
||||
if (!loginOk) { console.log('\n⛔ 登录失败，终止'); process.exit(1); }
|
||||
|
||||
const sessionOk = await testCreateSession();
|
||||
if (!sessionOk) { console.log('\n⛔ 会话创建失败，终止'); process.exit(1); }
|
||||
|
||||
await testDataProfile();
|
||||
|
||||
const goal = await testQLayer();
|
||||
if (!goal) { console.log('\n⚠️ Q 层失败，继续后续测试...'); }
|
||||
|
||||
const planOk = await testPLayer();
|
||||
if (!planOk) { console.log('\n⚠️ P 层失败，E/R 层将跳过'); }
|
||||
|
||||
const execOk = planOk ? await testELayer() : false;
|
||||
|
||||
if (execOk) {
|
||||
await testConclusionAPI();
|
||||
} else if (planOk) {
|
||||
console.log('\n⚠️ E 层失败，跳过结论 API 测试');
|
||||
}
|
||||
|
||||
await testSecondScenario();
|
||||
await testErrorClassification();
|
||||
|
||||
// 汇总
|
||||
console.log(`\n${'═'.repeat(60)}`);
|
||||
console.log(`📊 测试结果汇总：${passed} 通过 / ${failed} 失败 / ${skipped} 跳过 / ${passed + failed + skipped} 总计`);
|
||||
if (failed === 0) {
|
||||
console.log('🎉 全部通过！QPER 四层端到端验证成功。');
|
||||
} else {
|
||||
console.log(`⚠️ 有 ${failed} 个测试失败，请检查上方输出。`);
|
||||
}
|
||||
console.log(`\n📝 测试会话 ID: ${sessionId}`);
|
||||
console.log('═'.repeat(60));
|
||||
|
||||
process.exit(failed > 0 ? 1 : 0);
|
||||
}
|
||||
|
||||
main().catch(e => {
|
||||
console.error('💥 测试脚本异常:', e);
|
||||
process.exit(1);
|
||||
});
|
||||
Reference in New Issue
Block a user