refactor: 统一代码格式，调整多行代码换行风格
对多个文件进行代码格式化调整，将长行参数拆分为多行书写，提升代码可读性，包括： - 调整函数定义、调用的多行换行格式 - 优化列表、元组、字典的多行排版 - 新增README.md项目说明文档
This commit is contained in:
@@ -0,0 +1,309 @@
|
||||
# QMDSearch
|
||||
|
||||
面向 AI Agent 的分层信息检索服务，支持多层级知识库检索、向量语义搜索和结构化数据查询。基于 Docker 部署在 NAS 上，为 AI Agent 提供高效、精准的信息检索能力。
|
||||
|
||||
## 特性
|
||||
|
||||
- **分层检索架构**：L1 文档定位 → L2 章节大纲 → L3 小节定位 → chunk 原文，逐层收窄，精准召回
|
||||
- **Hybrid 检索**：dense 向量 + sparse 稀疏向量 RRF 融合，兼顾语义与关键词匹配
|
||||
- **文档三级总结**：Ollama 本地模型自动生成文档 L1/L2/L3 总结，支持知识分类
|
||||
- **多格式文件入库**：支持 `.txt` `.md` `.html` `.htm` `.pdf` `.docx`，PDF 扫描件自动 OCR 降级
|
||||
- **文本去重**：Sha256 文本去重，重复文档自动复用已有结果
|
||||
- **JWT 认证**：Token 鉴权，支持注册开关与默认管理员
|
||||
- **管理后台**：`/admin` 单页面，含概览、文档管理、检索测试台、类目管理
|
||||
- **Docker 部署**：开箱即用的 Docker Compose 编排，适合 NAS 环境
|
||||
|
||||
## 技术栈
|
||||
|
||||
| 组件 | 技术 |
|
||||
|------|------|
|
||||
| 语言 | Python 3.12+ |
|
||||
| Web 框架 | FastAPI + Uvicorn |
|
||||
| 向量数据库 | Qdrant（Docker） |
|
||||
| 缓存 | Redis（Docker） |
|
||||
| 嵌入模型 | OpenAI / Ollama 本地模型（可切换） |
|
||||
| 本地推理 | Ollama + qwen2.5:1.5b（文档总结） |
|
||||
| OCR | rapidocr-onnxruntime + pypdfium2（扫描件 PDF） |
|
||||
| 部署 | Docker Compose on NAS |
|
||||
|
||||
## 项目结构
|
||||
|
||||
```
|
||||
QMDSearch/
|
||||
├── app/ # 应用主目录
|
||||
│ ├── main.py # FastAPI 入口（含 /admin 管理页面）
|
||||
│ ├── config.py # 配置管理（pydantic-settings / 环境变量）
|
||||
│ ├── api/ # API 路由层
|
||||
│ │ ├── response.py # 统一响应格式
|
||||
│ │ └── v1/ # API v1
|
||||
│ │ ├── search.py # 检索接口
|
||||
│ │ ├── document.py # 文档入库/管理/上传
|
||||
│ │ ├── knowledge.py # 知识库接口
|
||||
│ │ └── auth.py # 认证接口
|
||||
│ ├── core/ # 核心业务逻辑
|
||||
│ │ ├── retriever.py # 分层检索引擎
|
||||
│ │ ├── query_parser.py # Query 解析与分类路由
|
||||
│ │ ├── embeddings.py # 向量嵌入
|
||||
│ │ ├── sparse.py # 稀疏向量编码
|
||||
│ │ ├── ranker.py # RRF 融合与结果截断
|
||||
│ │ ├── summarizer.py # 文档三级总结
|
||||
│ │ ├── headings.py # 原生标题树解析
|
||||
│ │ ├── classifier.py # 文档分类
|
||||
│ │ ├── chunker.py # 标题树感知 chunk 切分
|
||||
│ │ ├── ingestion.py # 文档入库流水线
|
||||
│ │ ├── ingest_tasks.py # 异步任务管理（含文本去重）
|
||||
│ │ ├── file_parser.py # 多格式文件解析（含 PDF OCR）
|
||||
│ │ └── auth.py # JWT 认证
|
||||
│ ├── models/ # Pydantic 数据模型
|
||||
│ ├── services/ # 外部服务客户端
|
||||
│ │ ├── qdrant.py # Qdrant 客户端
|
||||
│ │ ├── redis.py # Redis 缓存客户端
|
||||
│ │ └── ollama.py # Ollama 客户端
|
||||
│ └── static/ # 静态资源
|
||||
│ └── admin.html # 管理后台（单文件）
|
||||
├── tests/ # 测试
|
||||
├── scripts/ # 运维与评测脚本
|
||||
│ ├── eval/ # 回归评测
|
||||
│ └── smoke_live.py # 在线冒烟
|
||||
├── docker-compose.yml # Docker 编排
|
||||
├── Dockerfile # 应用镜像
|
||||
├── .env.example # 环境变量模板
|
||||
└── pyproject.toml # 项目配置
|
||||
```
|
||||
|
||||
## 分层检索架构
|
||||
|
||||
```
|
||||
Query → 解析路由 → L1（文档总结，候选文档集）
|
||||
↓
|
||||
L2（章节大纲，按 section 剪枝）
|
||||
↓
|
||||
L3（小节定位，两路 RRF 融合）
|
||||
↓
|
||||
Chunk（原文 hybrid 检索，dense + sparse）
|
||||
↓
|
||||
Top-K 结果
|
||||
```
|
||||
|
||||
### 检索链路详解
|
||||
|
||||
1. **Query 解析路由**：Ollama 小模型将 query 解析为结构化结果（rewrite、关键词、命中类目+置信度）。高置信且类目数不超上限时按类目过滤，否则全库兜底。
|
||||
2. **L1 - 文档定位层**：在文档 L1 总结集合中检索，产出候选文档集合。无命中时直接全库 chunk 兜底。
|
||||
3. **L2 - 章节大纲层**：在候选文档内检索 L2 章节大纲，按 section 剪枝定位。未命中的文档落入 L3 b 路。
|
||||
4. **L3 - 小节内容层**：两路查询（a：L2 命中文档按 section_path 过滤；b：其余文档仅按 doc_id 过滤）后 RRF 融合。
|
||||
5. **Chunk 层**：在 L3 收窄的范围内对原文 chunk 做 hybrid 检索（dense 向量 + sparse 稀疏向量 RRF 融合），返回最终 Top-K。
|
||||
|
||||
## 快速开始
|
||||
|
||||
### 前置要求
|
||||
|
||||
- Docker & Docker Compose
|
||||
- OpenAI API Key（或本地 Ollama 模型）
|
||||
|
||||
### 1. 克隆项目
|
||||
|
||||
```bash
|
||||
git clone <repo-url>
|
||||
cd QMDSearch
|
||||
```
|
||||
|
||||
### 2. 配置环境变量
|
||||
|
||||
```bash
|
||||
cp .env.example .env
|
||||
# 编辑 .env，至少配置 OPENAI_API_KEY
|
||||
```
|
||||
|
||||
### 3. 启动服务
|
||||
|
||||
```bash
|
||||
docker compose up -d
|
||||
```
|
||||
|
||||
### 4. 拉取 Ollama 模型（首次启动后）
|
||||
|
||||
```bash
|
||||
docker exec qmdsearch-ollama ollama pull qwen2.5:1.5b
|
||||
```
|
||||
|
||||
### 5. 验证服务
|
||||
|
||||
```bash
|
||||
curl http://localhost:8000/api/v1/health
|
||||
# {"status": "ok"}
|
||||
```
|
||||
|
||||
管理后台：浏览器访问 `http://localhost:8000/admin`
|
||||
|
||||
## API 文档
|
||||
|
||||
| 方法 | 路径 | 说明 | 认证 |
|
||||
|------|------|------|------|
|
||||
| `GET` | `/api/v1/health` | 健康检查 | 否 |
|
||||
| `POST` | `/api/v1/auth/register` | 用户注册 | 否（可关闭） |
|
||||
| `POST` | `/api/v1/auth/login` | 用户登录，返回 JWT token | 否 |
|
||||
| `GET` | `/api/v1/auth/me` | 获取当前用户信息 | 是 |
|
||||
| `POST` | `/api/v1/search` | 分层检索 | 是 |
|
||||
| `POST` | `/api/v1/documents` | 文档入库（JSON 文本，202 异步入库） | 是 |
|
||||
| `POST` | `/api/v1/documents/upload` | 文件上传入库（multipart，202 异步） | 是 |
|
||||
| `GET` | `/api/v1/documents/tasks/{task_id}` | 入库任务状态查询 | 是 |
|
||||
| `GET` | `/api/v1/documents` | 文档列表（分页） | 是 |
|
||||
| `GET` | `/api/v1/documents/{doc_id}` | 文档详情 | 是 |
|
||||
| `DELETE` | `/api/v1/documents/{doc_id}` | 删除文档（幂等） | 是 |
|
||||
| `GET` | `/api/v1/knowledge/categories` | 知识分类类目集 | 是 |
|
||||
| `GET` | `/api/v1/knowledge/stats` | 统计（四层点数 + 类目分布） | 是 |
|
||||
| `GET` | `/admin` | 管理后台 | 否 |
|
||||
|
||||
### 统一响应格式
|
||||
|
||||
```json
|
||||
{
|
||||
"code": 0,
|
||||
"data": { ... },
|
||||
"message": "ok"
|
||||
}
|
||||
```
|
||||
|
||||
错误码：`0` 成功，`1xxx` 客户端错误，`2xxx` 服务端错误。
|
||||
|
||||
### 检索示例
|
||||
|
||||
```bash
|
||||
curl -X POST http://localhost:8000/api/v1/search \
|
||||
-H "Content-Type: application/json" \
|
||||
-H "Authorization: Bearer <token>" \
|
||||
-d '{"query": "如何配置 Redis 缓存", "top_k": 5}'
|
||||
```
|
||||
|
||||
### 文件上传示例
|
||||
|
||||
```bash
|
||||
curl -X POST http://localhost:8000/api/v1/documents/upload \
|
||||
-H "Authorization: Bearer <token>" \
|
||||
-F "file=@document.pdf"
|
||||
```
|
||||
|
||||
## 文档入库流程
|
||||
|
||||
```
|
||||
文件上传 → 文本提取 → 三级总结（Ollama）→ 分类判定（L1 总结）→ 向量化 → 写入 Qdrant
|
||||
↓
|
||||
不足三级 → 2.5 级回退
|
||||
```
|
||||
|
||||
### 三级总结
|
||||
|
||||
| 层级 | 名称 | 说明 |
|
||||
|------|------|------|
|
||||
| L1 | 总结 | 对整篇文档的一句话高度概括 |
|
||||
| L2 | 大纲 | 文档主要章节和关键主题 |
|
||||
| L3 | 内容大纲 | 每个章节的详细内容摘要 |
|
||||
|
||||
**2.5 级回退**：短文本（< 500 字符）自动降级为二级总结，跳过大纲层。
|
||||
|
||||
### 文件格式支持
|
||||
|
||||
| 格式 | 解析方式 |
|
||||
|------|---------|
|
||||
| `.txt` `.md` | UTF-8 解码 |
|
||||
| `.html` `.htm` | HTML 标签剥离 |
|
||||
| `.pdf` | pypdf 提取文本层；扫描件自动 OCR 降级 |
|
||||
| `.docx` | python-docx 段落提取 |
|
||||
|
||||
### PDF OCR 降级
|
||||
|
||||
图片型/扫描件 PDF 在文本层为空时，自动触发 OCR：
|
||||
1. pypdfium2 渲染每页为图片
|
||||
2. rapidocr-onnxruntime 识别文字
|
||||
3. 拼接返回结果
|
||||
|
||||
可通过环境变量 `PDF_OCR_ENABLED=false` 关闭，或通过 `PDF_OCR_MAX_PAGES` / `PDF_OCR_DPI` 控制行为。
|
||||
|
||||
### 文本去重
|
||||
|
||||
入库时自动计算 `sha256(doc.text)`，相同文本重复提交不重跑流水线，直接复用已有文档 ID，结果中 `deduplicated=true` 标记。
|
||||
|
||||
## 配置项
|
||||
|
||||
所有配置通过环境变量注入，完整列表见 `.env.example`。主要配置项：
|
||||
|
||||
| 变量 | 说明 | 默认值 |
|
||||
|------|------|--------|
|
||||
| `EMBEDDING_PROVIDER` | 嵌入模型提供商（openai / local） | `openai` |
|
||||
| `OPENAI_API_KEY` | OpenAI API Key | - |
|
||||
| `EMBEDDING_MODEL` | 嵌入模型名称 | `text-embedding-3-small` |
|
||||
| `OLLAMA_MODEL` | Ollama 本地模型 | `qwen2.5:1.5b` |
|
||||
| `RETRIEVAL_TOP_K` | 检索召回数 | `20` |
|
||||
| `RETRIEVAL_FINAL_K` | 最终返回数 | `5` |
|
||||
| `INGEST_MAX_CONCURRENCY` | 入库并发上限 | `2` |
|
||||
| `JWT_SECRET_KEY` | JWT 签名密钥（生产必填） | 自动生成（仅开发） |
|
||||
| `AUTH_REGISTER_ENABLED` | 是否开放注册 | `true` |
|
||||
| `UPLOAD_MAX_SIZE_MB` | 上传文件大小上限 | `20` |
|
||||
| `PDF_OCR_ENABLED` | 是否启用 PDF OCR | `true` |
|
||||
| `PDF_OCR_MAX_PAGES` | OCR 最大页数 | `30` |
|
||||
| `PDF_OCR_DPI` | OCR 渲染 DPI | `200` |
|
||||
|
||||
## 开发
|
||||
|
||||
### 本地开发
|
||||
|
||||
```bash
|
||||
# 安装依赖
|
||||
uv sync --extra dev
|
||||
|
||||
# 启动开发服务器
|
||||
uv run uvicorn app.main:app --reload --host 0.0.0.0 --port 8000
|
||||
|
||||
# 运行测试
|
||||
uv run pytest
|
||||
|
||||
# 代码检查
|
||||
uv run ruff check app/ tests/
|
||||
```
|
||||
|
||||
### 构建镜像
|
||||
|
||||
```bash
|
||||
docker compose build
|
||||
```
|
||||
|
||||
### 目录结构约定
|
||||
|
||||
- 类型注解必须（Python 3.12+ 语法）
|
||||
- 异步优先（async/await）
|
||||
- 配置通过环境变量注入（pydantic-settings）
|
||||
- 日志使用 structlog 结构化日志
|
||||
- 测试使用 pytest + pytest-asyncio
|
||||
|
||||
## 部署
|
||||
|
||||
### NAS 部署
|
||||
|
||||
```bash
|
||||
# 1. 创建数据目录
|
||||
mkdir -p /path/to/nas/data/{qdrant,redis,ollama}
|
||||
|
||||
# 2. 配置 .env，设置 NAS_DATA_DIR 和 API Key
|
||||
NAS_DATA_DIR=/path/to/nas/data
|
||||
OPENAI_API_KEY=sk-xxx
|
||||
|
||||
# 3. 启动
|
||||
docker compose up -d
|
||||
|
||||
# 4. 查看日志
|
||||
docker compose logs -f app
|
||||
```
|
||||
|
||||
### 端口映射
|
||||
|
||||
| 服务 | 端口 |
|
||||
|------|------|
|
||||
| API | 8000 |
|
||||
| Qdrant | 6333 |
|
||||
| Qdrant Dashboard | 6334 |
|
||||
| Redis | 6379 |
|
||||
| Ollama | 11434 |
|
||||
|
||||
## License
|
||||
|
||||
MIT
|
||||
+44
-13
@@ -57,30 +57,44 @@ def _get_task_manager() -> IngestTaskManager:
|
||||
except Exception:
|
||||
logger.warning("Redis 缓存不可用，入库任务状态仅保留在内存", exc_info=True)
|
||||
redis = None
|
||||
_task_manager = IngestTaskManager(ingester=_get_ingester(), redis=redis, settings=Settings())
|
||||
_task_manager = IngestTaskManager(
|
||||
ingester=_get_ingester(), redis=redis, settings=Settings()
|
||||
)
|
||||
return _task_manager
|
||||
|
||||
|
||||
def _allowed_extensions() -> set[str]:
|
||||
"""解析 settings.upload_allowed_extensions 逗号分隔字符串为扩展名集合（全小写、含点号）"""
|
||||
return {ext.strip().lower() for ext in settings.upload_allowed_extensions.split(",") if ext.strip()}
|
||||
return {
|
||||
ext.strip().lower()
|
||||
for ext in settings.upload_allowed_extensions.split(",")
|
||||
if ext.strip()
|
||||
}
|
||||
|
||||
|
||||
@router.post("/documents")
|
||||
async def ingest_document(doc: DocumentInput, user: AuthUser = Depends(get_current_user)) -> JSONResponse:
|
||||
async def ingest_document(
|
||||
doc: DocumentInput, user: AuthUser = Depends(get_current_user)
|
||||
) -> JSONResponse:
|
||||
"""文档入库入口：登记异步任务并返回 202 + task_id，入库结果经任务查询端点获取"""
|
||||
if not doc.text.strip():
|
||||
raise ApiError(1001, "文档内容不能为空")
|
||||
task_id = await _get_task_manager().submit(doc)
|
||||
return JSONResponse(status_code=202, content=ok({"task_id": task_id, "status": "pending"}))
|
||||
return JSONResponse(
|
||||
status_code=202, content=ok({"task_id": task_id, "status": "pending"})
|
||||
)
|
||||
|
||||
|
||||
@router.post("/documents/upload")
|
||||
async def upload_document(
|
||||
file: UploadFile = File(..., description="上传的文件（.txt/.md/.html/.htm/.pdf/.docx）"),
|
||||
file: UploadFile = File(
|
||||
..., description="上传的文件（.txt/.md/.html/.htm/.pdf/.docx）"
|
||||
),
|
||||
title: str = Form(default="", description="可选标题，默认取原文件名去扩展"),
|
||||
source: str = Form(default="", description="可选来源标识，默认 file:{原文件名}"),
|
||||
metadata: str = Form(default="", description='可选元数据 JSON 字符串，如 \'{"author":"x"}\''),
|
||||
metadata: str = Form(
|
||||
default="", description='可选元数据 JSON 字符串，如 \'{"author":"x"}\''
|
||||
),
|
||||
user: AuthUser = Depends(get_current_user),
|
||||
) -> JSONResponse:
|
||||
"""文件上传入库入口：校验 → 提取文本 → 落盘 → 提交异步入库流水线
|
||||
@@ -129,9 +143,16 @@ async def upload_document(
|
||||
"original_size_bytes": str(len(content)),
|
||||
}
|
||||
)
|
||||
logger.info("上传文件已落盘", doc_id=doc_id, saved_path=saved_path, size=len(content))
|
||||
logger.info(
|
||||
"上传文件已落盘", doc_id=doc_id, saved_path=saved_path, size=len(content)
|
||||
)
|
||||
except Exception:
|
||||
logger.warning("上传文件落盘失败，仅做文本入库", doc_id=doc_id, filename=original_filename, exc_info=True)
|
||||
logger.warning(
|
||||
"上传文件落盘失败，仅做文本入库",
|
||||
doc_id=doc_id,
|
||||
filename=original_filename,
|
||||
exc_info=True,
|
||||
)
|
||||
|
||||
# 5. 合并用户传入的 metadata（落盘元数据优先级更高，不与用户键冲突）
|
||||
if metadata.strip():
|
||||
@@ -150,7 +171,9 @@ async def upload_document(
|
||||
source = f"file:{original_filename}"
|
||||
|
||||
# 7. 提交入库流水线
|
||||
doc_input = DocumentInput(text=text, title=title, source=source, metadata=metadata_dict)
|
||||
doc_input = DocumentInput(
|
||||
text=text, title=title, source=source, metadata=metadata_dict
|
||||
)
|
||||
task_id = await _get_task_manager().submit(doc_input)
|
||||
|
||||
return JSONResponse(
|
||||
@@ -160,7 +183,9 @@ async def upload_document(
|
||||
|
||||
|
||||
@router.get("/documents/tasks/{task_id}")
|
||||
async def get_ingest_task(task_id: str, user: AuthUser = Depends(get_current_user)) -> dict[str, Any]:
|
||||
async def get_ingest_task(
|
||||
task_id: str, user: AuthUser = Depends(get_current_user)
|
||||
) -> dict[str, Any]:
|
||||
"""查询入库任务状态：含 task_id/status/created_at/updated_at，done 附 result，failed 附 error"""
|
||||
task = await _get_task_manager().get(task_id)
|
||||
if task is None:
|
||||
@@ -184,7 +209,9 @@ async def list_documents(
|
||||
|
||||
|
||||
@router.get("/documents/{doc_id}")
|
||||
async def get_document(doc_id: str, user: AuthUser = Depends(get_current_user)) -> dict[str, Any]:
|
||||
async def get_document(
|
||||
doc_id: str, user: AuthUser = Depends(get_current_user)
|
||||
) -> dict[str, Any]:
|
||||
"""获取文档详情：L1 记录 + L2/L3 节点 + chunks 数量"""
|
||||
try:
|
||||
detail = await _get_qdrant().get_doc_detail(doc_id)
|
||||
@@ -197,11 +224,15 @@ async def get_document(doc_id: str, user: AuthUser = Depends(get_current_user))
|
||||
|
||||
|
||||
@router.delete("/documents/{doc_id}")
|
||||
async def delete_document(doc_id: str, user: AuthUser = Depends(require_admin)) -> dict[str, Any]:
|
||||
async def delete_document(
|
||||
doc_id: str, user: AuthUser = Depends(require_admin)
|
||||
) -> dict[str, Any]:
|
||||
"""删除文档：四层集合中该 doc_id 的所有点；幂等，不存在也返回成功（删除数全 0）"""
|
||||
try:
|
||||
deleted = await _get_qdrant().delete_by_doc_id(doc_id)
|
||||
except Exception as exc:
|
||||
logger.error("文档删除失败", doc_id=doc_id, error=str(exc))
|
||||
raise ApiError(2000, f"文档删除失败: {exc}") from exc
|
||||
return ok({"doc_id": doc_id, "deleted": deleted, "deleted_total": sum(deleted.values())})
|
||||
return ok(
|
||||
{"doc_id": doc_id, "deleted": deleted, "deleted_total": sum(deleted.values())}
|
||||
)
|
||||
|
||||
+6
-2
@@ -65,10 +65,14 @@ class Settings(BaseSettings):
|
||||
# 文件上传
|
||||
upload_dir: str = "./uploads" # 原始文件保存目录（相对路径以工作目录为基）
|
||||
upload_max_size_mb: int = 20 # 单文件大小上限（MB）
|
||||
upload_allowed_extensions: str = ".txt,.md,.html,.htm,.pdf,.docx" # 允许上传的扩展名（逗号分隔）
|
||||
upload_allowed_extensions: str = (
|
||||
".txt,.md,.html,.htm,.pdf,.docx" # 允许上传的扩展名（逗号分隔）
|
||||
)
|
||||
|
||||
# PDF OCR（图片型/扫描件降级，pypdf extract_text 为空时触发）
|
||||
pdf_ocr_enabled: bool = True # 是否启用 OCR 降级（关闭则扫描件按"无法提取文本"拒绝入库）
|
||||
pdf_ocr_enabled: bool = (
|
||||
True # 是否启用 OCR 降级（关闭则扫描件按"无法提取文本"拒绝入库）
|
||||
)
|
||||
pdf_ocr_max_pages: int = 30 # 单文件 OCR 页数上限，超过仅前 N 页
|
||||
pdf_ocr_dpi: int = 200 # 渲染 DPI（越高越准但越慢，72~300 合理）
|
||||
|
||||
|
||||
@@ -102,7 +102,9 @@ def _parse_pdf(content: bytes) -> str:
|
||||
if not settings.pdf_ocr_enabled:
|
||||
return ""
|
||||
|
||||
ocr_text = _ocr_pdf(content, max_pages=settings.pdf_ocr_max_pages, dpi=settings.pdf_ocr_dpi)
|
||||
ocr_text = _ocr_pdf(
|
||||
content, max_pages=settings.pdf_ocr_max_pages, dpi=settings.pdf_ocr_dpi
|
||||
)
|
||||
return ocr_text
|
||||
|
||||
|
||||
@@ -143,7 +145,9 @@ def _ocr_pdf(content: bytes, max_pages: int, dpi: int) -> str:
|
||||
result, _ = _ocr_engine(pil_image)
|
||||
if result:
|
||||
# result: [[box, text, score], ...]，按行拼接
|
||||
lines = [item[1] for item in result if item and len(item) >= 2 and item[1]]
|
||||
lines = [
|
||||
item[1] for item in result if item and len(item) >= 2 and item[1]
|
||||
]
|
||||
if lines:
|
||||
page_texts.append("\n".join(lines))
|
||||
pdf.close()
|
||||
|
||||
+44
-12
@@ -43,7 +43,9 @@ class IngestTaskStatus(StrEnum):
|
||||
|
||||
|
||||
# 终态集合
|
||||
TERMINAL_STATUSES: frozenset[str] = frozenset({IngestTaskStatus.DONE, IngestTaskStatus.FAILED})
|
||||
TERMINAL_STATUSES: frozenset[str] = frozenset(
|
||||
{IngestTaskStatus.DONE, IngestTaskStatus.FAILED}
|
||||
)
|
||||
|
||||
|
||||
def _utc_now_iso() -> str:
|
||||
@@ -54,7 +56,9 @@ def _utc_now_iso() -> str:
|
||||
class IngestTaskManager:
|
||||
"""入库异步任务管理器：登记、后台执行、状态查询与 Redis 持久镜像"""
|
||||
|
||||
def __init__(self, ingester: Ingester, redis: RedisCache | None, settings: Settings) -> None:
|
||||
def __init__(
|
||||
self, ingester: Ingester, redis: RedisCache | None, settings: Settings
|
||||
) -> None:
|
||||
self._ingester = ingester
|
||||
self._redis = redis
|
||||
self._settings = settings
|
||||
@@ -123,7 +127,11 @@ class IngestTaskManager:
|
||||
try:
|
||||
return await self._redis.get_json(f"{REDIS_KEY_PREFIX}{task_id}")
|
||||
except Exception:
|
||||
logger.warning("入库任务状态读取 Redis 失败，降级为未命中", task_id=task_id, exc_info=True)
|
||||
logger.warning(
|
||||
"入库任务状态读取 Redis 失败，降级为未命中",
|
||||
task_id=task_id,
|
||||
exc_info=True,
|
||||
)
|
||||
return None
|
||||
|
||||
async def wait_done(self, task_id: str, timeout: float = 30.0) -> dict[str, Any]:
|
||||
@@ -147,7 +155,9 @@ class IngestTaskManager:
|
||||
"""后台执行入库：并发限流 + 阶段状态推进 + 结果/错误落账 + 去重记录写入"""
|
||||
async with self._semaphore:
|
||||
try:
|
||||
result = await self._ingester.ingest(doc, progress_cb=lambda stage: self._on_progress(task_id, stage))
|
||||
result = await self._ingester.ingest(
|
||||
doc, progress_cb=lambda stage: self._on_progress(task_id, stage)
|
||||
)
|
||||
except IngestionError as exc:
|
||||
# 入库已知失败：透传阶段与已产出的部分总结
|
||||
self._finish_failed(
|
||||
@@ -155,11 +165,18 @@ class IngestTaskManager:
|
||||
{
|
||||
"stage": exc.stage,
|
||||
"message": str(exc),
|
||||
"partial_summary": exc.summary.model_dump(mode="json") if exc.summary is not None else None,
|
||||
"partial_summary": (
|
||||
exc.summary.model_dump(mode="json")
|
||||
if exc.summary is not None
|
||||
else None
|
||||
),
|
||||
},
|
||||
)
|
||||
except Exception as exc:
|
||||
self._finish_failed(task_id, {"stage": "unknown", "message": str(exc), "partial_summary": None})
|
||||
self._finish_failed(
|
||||
task_id,
|
||||
{"stage": "unknown", "message": str(exc), "partial_summary": None},
|
||||
)
|
||||
else:
|
||||
result_dict = result.model_dump(mode="json")
|
||||
self._tasks[task_id].update(
|
||||
@@ -178,7 +195,9 @@ class IngestTaskManager:
|
||||
try:
|
||||
return await self._redis.get_json(f"{DEDUP_KEY_PREFIX}{text_hash}")
|
||||
except Exception:
|
||||
logger.warning("去重记录查询失败，降级为未命中", text_hash=text_hash, exc_info=True)
|
||||
logger.warning(
|
||||
"去重记录查询失败，降级为未命中", text_hash=text_hash, exc_info=True
|
||||
)
|
||||
return None
|
||||
|
||||
async def _record_dedup(self, text_hash: str, result_dict: dict[str, Any]) -> None:
|
||||
@@ -196,9 +215,16 @@ class IngestTaskManager:
|
||||
|
||||
def _finish_failed(self, task_id: str, error: dict[str, Any]) -> None:
|
||||
"""将任务置为 failed 并记录错误信息"""
|
||||
self._tasks[task_id].update(status=IngestTaskStatus.FAILED, updated_at=_utc_now_iso(), error=error)
|
||||
self._tasks[task_id].update(
|
||||
status=IngestTaskStatus.FAILED, updated_at=_utc_now_iso(), error=error
|
||||
)
|
||||
self._schedule_mirror(task_id)
|
||||
logger.error("入库任务失败", task_id=task_id, stage=error["stage"], error=error["message"])
|
||||
logger.error(
|
||||
"入库任务失败",
|
||||
task_id=task_id,
|
||||
stage=error["stage"],
|
||||
error=error["message"],
|
||||
)
|
||||
|
||||
def _on_progress(self, task_id: str, stage: str) -> None:
|
||||
"""Ingester 阶段回调：推进任务状态并同步镜像（同步函数，供 progress_cb 使用）"""
|
||||
@@ -213,7 +239,9 @@ class IngestTaskManager:
|
||||
"""将当前任务状态快照异步镜像到 Redis（同步上下文也可调用）"""
|
||||
if self._redis is None:
|
||||
return
|
||||
mirror = asyncio.create_task(self._mirror_to_redis(task_id, dict(self._tasks[task_id])))
|
||||
mirror = asyncio.create_task(
|
||||
self._mirror_to_redis(task_id, dict(self._tasks[task_id]))
|
||||
)
|
||||
self._mirror_tasks.add(mirror)
|
||||
mirror.add_done_callback(self._mirror_tasks.discard)
|
||||
|
||||
@@ -225,6 +253,10 @@ class IngestTaskManager:
|
||||
else self._settings.ingest_task_ttl_done
|
||||
)
|
||||
try:
|
||||
await self._redis.set_json(f"{REDIS_KEY_PREFIX}{task_id}", snapshot, ttl=ttl)
|
||||
await self._redis.set_json(
|
||||
f"{REDIS_KEY_PREFIX}{task_id}", snapshot, ttl=ttl
|
||||
)
|
||||
except Exception:
|
||||
logger.warning("入库任务状态镜像 Redis 失败", task_id=task_id, exc_info=True)
|
||||
logger.warning(
|
||||
"入库任务状态镜像 Redis 失败", task_id=task_id, exc_info=True
|
||||
)
|
||||
|
||||
+3
-1
@@ -62,7 +62,9 @@ async def api_error_handler(request: Request, exc: ApiError) -> JSONResponse:
|
||||
|
||||
|
||||
@app.exception_handler(RequestValidationError)
|
||||
async def validation_error_handler(request: Request, exc: RequestValidationError) -> JSONResponse:
|
||||
async def validation_error_handler(
|
||||
request: Request, exc: RequestValidationError
|
||||
) -> JSONResponse:
|
||||
"""请求参数校验失败 → code 1001"""
|
||||
errors = exc.errors()
|
||||
message = f"请求参数校验失败: {errors[0]['msg']}" if errors else "请求参数校验失败"
|
||||
|
||||
@@ -16,7 +16,9 @@ class DocumentSummary(BaseModel):
|
||||
"""文档三级总结结果"""
|
||||
|
||||
l1_summary: str = Field(description="L1 总结：一句话高度概括")
|
||||
l2_outline: str | None = Field(default=None, description="L2 大纲：主要章节和关键主题")
|
||||
l2_outline: str | None = Field(
|
||||
default=None, description="L2 大纲：主要章节和关键主题"
|
||||
)
|
||||
l3_content_outline: str = Field(description="L3/L2.5 内容大纲：详细内容摘要")
|
||||
level: SummaryLevel = Field(description="实际使用的总结层级")
|
||||
|
||||
@@ -49,4 +51,7 @@ class IngestionResult(BaseModel):
|
||||
chunks_count: int = Field(default=0, description="写入的 chunk 数量")
|
||||
tags: list[str] = Field(default_factory=list, description="附加分类标签")
|
||||
category_confidence: float = Field(default=0.0, description="主类目分类置信度")
|
||||
deduplicated: bool = Field(default=False, description="是否命中去重复用旧文档（True 时 document_id 为既有文档 ID）")
|
||||
deduplicated: bool = Field(
|
||||
default=False,
|
||||
description="是否命中去重复用旧文档（True 时 document_id 为既有文档 ID）",
|
||||
)
|
||||
|
||||
@@ -34,7 +34,9 @@ def client(monkeypatch: pytest.MonkeyPatch, tmp_path: Path) -> Iterator[TestClie
|
||||
return None
|
||||
|
||||
monkeypatch.setattr(QdrantService, "ensure_collections", _noop_ensure_collections)
|
||||
monkeypatch.setattr(document_module.settings, "upload_dir", str(tmp_path / "uploads"))
|
||||
monkeypatch.setattr(
|
||||
document_module.settings, "upload_dir", str(tmp_path / "uploads")
|
||||
)
|
||||
with TestClient(app) as test_client:
|
||||
yield test_client
|
||||
|
||||
@@ -45,7 +47,13 @@ def _inject_manager(monkeypatch: pytest.MonkeyPatch, manager: FakeManager) -> No
|
||||
|
||||
def _make_minimal_pdf(text: str = "Hello PDF World") -> bytes:
|
||||
content_stream = f"BT /F1 24 Tf 100 700 Td ({text}) Tj ET".encode("latin-1")
|
||||
content_obj = b"<< /Length " + str(len(content_stream)).encode() + b" >>\nstream\n" + content_stream + b"\nendstream"
|
||||
content_obj = (
|
||||
b"<< /Length "
|
||||
+ str(len(content_stream)).encode()
|
||||
+ b" >>\nstream\n"
|
||||
+ content_stream
|
||||
+ b"\nendstream"
|
||||
)
|
||||
return (
|
||||
b"%PDF-1.0\n"
|
||||
b"1 0 obj\n<< /Type /Catalog /Pages 2 0 R >>\nendobj\n"
|
||||
@@ -161,7 +169,13 @@ def test_upload_docx_extracts_text(
|
||||
content = _make_docx(["第一段落", "第二段落"])
|
||||
resp = client.post(
|
||||
"/api/v1/documents/upload",
|
||||
files={"file": ("doc.docx", content, "application/vnd.openxmlformats-officedocument.wordprocessingml.document")},
|
||||
files={
|
||||
"file": (
|
||||
"doc.docx",
|
||||
content,
|
||||
"application/vnd.openxmlformats-officedocument.wordprocessingml.document",
|
||||
)
|
||||
},
|
||||
)
|
||||
|
||||
assert resp.status_code == 202
|
||||
|
||||
@@ -14,7 +14,13 @@ from app.core.file_parser import parse_file, supported_extensions
|
||||
def _make_minimal_pdf(text: str = "Hello PDF World") -> bytes:
|
||||
"""构造一个含一页文本的最小 PDF（pypdf 可读出文本）"""
|
||||
content_stream = f"BT /F1 24 Tf 100 700 Td ({text}) Tj ET".encode("latin-1")
|
||||
content_obj = b"<< /Length " + str(len(content_stream)).encode() + b" >>\nstream\n" + content_stream + b"\nendstream"
|
||||
content_obj = (
|
||||
b"<< /Length "
|
||||
+ str(len(content_stream)).encode()
|
||||
+ b" >>\nstream\n"
|
||||
+ content_stream
|
||||
+ b"\nendstream"
|
||||
)
|
||||
return (
|
||||
b"%PDF-1.0\n"
|
||||
b"1 0 obj\n<< /Type /Catalog /Pages 2 0 R >>\nendobj\n"
|
||||
@@ -36,6 +42,7 @@ def _make_minimal_pdf(text: str = "Hello PDF World") -> bytes:
|
||||
|
||||
def _make_docx(text_lines: list[str]) -> bytes:
|
||||
from docx import Document # type: ignore[import-untyped]
|
||||
|
||||
document = Document()
|
||||
for line in text_lines:
|
||||
document.add_paragraph(line)
|
||||
@@ -222,7 +229,9 @@ def _patch_pdf_ocr_deps(monkeypatch: pytest.MonkeyPatch) -> None:
|
||||
monkeypatch.setitem(sys.modules, "rapidocr_onnxruntime", fake_module)
|
||||
|
||||
|
||||
def test_parse_pdf_ocr_fallback_when_text_layer_empty(monkeypatch: pytest.MonkeyPatch) -> None:
|
||||
def test_parse_pdf_ocr_fallback_when_text_layer_empty(
|
||||
monkeypatch: pytest.MonkeyPatch,
|
||||
) -> None:
|
||||
"""扫描件 PDF（文本层全空）触发 OCR 降级，返回识别文本"""
|
||||
_patch_pdf_ocr_deps(monkeypatch)
|
||||
monkeypatch.setattr(settings, "pdf_ocr_enabled", True)
|
||||
@@ -256,7 +265,9 @@ def test_parse_pdf_ocr_respects_max_pages(monkeypatch: pytest.MonkeyPatch) -> No
|
||||
assert "OCR文本第2页" not in result
|
||||
|
||||
|
||||
def test_parse_pdf_ocr_returns_empty_when_dependency_unavailable(monkeypatch: pytest.MonkeyPatch) -> None:
|
||||
def test_parse_pdf_ocr_returns_empty_when_dependency_unavailable(
|
||||
monkeypatch: pytest.MonkeyPatch,
|
||||
) -> None:
|
||||
"""rapidocr 导入失败：降级返回空文本，且把 _ocr_unavailable 置 True 避免重试"""
|
||||
monkeypatch.setattr("pypdf.PdfReader", _FakePdfReader)
|
||||
# 故意让 rapidocr_onnxruntime 提供一个非类的 RapidOCR，构造时抛错
|
||||
@@ -273,7 +284,9 @@ def test_parse_pdf_ocr_returns_empty_when_dependency_unavailable(monkeypatch: py
|
||||
assert fp_module._ocr_unavailable is True
|
||||
|
||||
|
||||
def test_parse_pdf_ocr_runtime_exception_falls_back_to_empty(monkeypatch: pytest.MonkeyPatch) -> None:
|
||||
def test_parse_pdf_ocr_runtime_exception_falls_back_to_empty(
|
||||
monkeypatch: pytest.MonkeyPatch,
|
||||
) -> None:
|
||||
"""OCR 运行时抛错：仅告警，降级返回空文本（不抛出 ValueError）"""
|
||||
monkeypatch.setattr("pypdf.PdfReader", _FakePdfReader)
|
||||
|
||||
@@ -291,7 +304,9 @@ def test_parse_pdf_ocr_runtime_exception_falls_back_to_empty(monkeypatch: pytest
|
||||
assert parse_file("scan.pdf", b"fake pdf bytes") == ""
|
||||
|
||||
|
||||
def test_parse_pdf_text_layer_present_skips_ocr(monkeypatch: pytest.MonkeyPatch) -> None:
|
||||
def test_parse_pdf_text_layer_present_skips_ocr(
|
||||
monkeypatch: pytest.MonkeyPatch,
|
||||
) -> None:
|
||||
"""文本层非空：直接返回文本，OCR 引擎不会被实例化"""
|
||||
call_count = 0
|
||||
|
||||
|
||||
@@ -7,15 +7,28 @@ from datetime import datetime
|
||||
from typing import Any
|
||||
|
||||
from app.config import Settings
|
||||
from app.core.ingest_tasks import DEDUP_KEY_PREFIX, REDIS_KEY_PREFIX, IngestTaskManager, IngestTaskStatus
|
||||
from app.core.ingest_tasks import (
|
||||
DEDUP_KEY_PREFIX,
|
||||
REDIS_KEY_PREFIX,
|
||||
IngestTaskManager,
|
||||
IngestTaskStatus,
|
||||
)
|
||||
from app.core.ingestion import IngestionError
|
||||
from app.models.document import DocumentInput, DocumentSummary, IngestionResult, SummaryLevel
|
||||
from app.models.document import (
|
||||
DocumentInput,
|
||||
DocumentSummary,
|
||||
IngestionResult,
|
||||
SummaryLevel,
|
||||
)
|
||||
|
||||
|
||||
def make_summary() -> DocumentSummary:
|
||||
"""构造固定的三级总结"""
|
||||
return DocumentSummary(
|
||||
l1_summary="一句话总结", l2_outline=None, l3_content_outline="内容大纲", level=SummaryLevel.L3
|
||||
l1_summary="一句话总结",
|
||||
l2_outline=None,
|
||||
l3_content_outline="内容大纲",
|
||||
level=SummaryLevel.L3,
|
||||
)
|
||||
|
||||
|
||||
@@ -42,7 +55,9 @@ class FakeIngester:
|
||||
self.gate: asyncio.Event | None = None
|
||||
self.started = asyncio.Event()
|
||||
|
||||
async def ingest(self, doc: DocumentInput, progress_cb: Callable[[str], None] | None = None) -> IngestionResult:
|
||||
async def ingest(
|
||||
self, doc: DocumentInput, progress_cb: Callable[[str], None] | None = None
|
||||
) -> IngestionResult:
|
||||
self.calls.append(doc)
|
||||
self.started.set()
|
||||
if progress_cb is not None:
|
||||
@@ -64,7 +79,9 @@ class FakeRedis:
|
||||
self.writes: list[tuple[str, dict[str, Any], int | None]] = []
|
||||
self.store: dict[str, dict[str, Any]] = {}
|
||||
|
||||
async def set_json(self, key: str, value: dict[str, Any], ttl: int | None = None) -> bool:
|
||||
async def set_json(
|
||||
self, key: str, value: dict[str, Any], ttl: int | None = None
|
||||
) -> bool:
|
||||
self.writes.append((key, value, ttl))
|
||||
if self.fail_writes:
|
||||
raise RuntimeError("redis down")
|
||||
@@ -105,8 +122,12 @@ async def test_submit_returns_immediately_and_completes() -> None:
|
||||
|
||||
async def test_ingestion_error_marks_failed_with_stage_and_partial_summary() -> None:
|
||||
"""IngestionError：任务 failed，error.stage 透传，partial_summary 保留"""
|
||||
summary = DocumentSummary(l1_summary="L1", l2_outline=None, l3_content_outline="L3", level=SummaryLevel.L3)
|
||||
ingester = FakeIngester(error=IngestionError("classify", "分类判定失败: boom", summary=summary))
|
||||
summary = DocumentSummary(
|
||||
l1_summary="L1", l2_outline=None, l3_content_outline="L3", level=SummaryLevel.L3
|
||||
)
|
||||
ingester = FakeIngester(
|
||||
error=IngestionError("classify", "分类判定失败: boom", summary=summary)
|
||||
)
|
||||
manager = IngestTaskManager(ingester, FakeRedis(), Settings())
|
||||
|
||||
task_id = await manager.submit(DocumentInput(text="正文"))
|
||||
@@ -179,10 +200,14 @@ async def test_redis_mirror_ttl_done_and_failed() -> None:
|
||||
assert done_writes[-1][0]["status"] == IngestTaskStatus.DONE
|
||||
|
||||
redis2 = FakeRedis()
|
||||
failing_manager = IngestTaskManager(FakeIngester(error=RuntimeError("boom")), redis2, settings)
|
||||
failing_manager = IngestTaskManager(
|
||||
FakeIngester(error=RuntimeError("boom")), redis2, settings
|
||||
)
|
||||
failed_id = await failing_manager.submit(DocumentInput(text="y"))
|
||||
await failing_manager.wait_done(failed_id, timeout=5)
|
||||
failed_writes = [(v, ttl) for key, v, ttl in redis2.writes if key.endswith(failed_id)]
|
||||
failed_writes = [
|
||||
(v, ttl) for key, v, ttl in redis2.writes if key.endswith(failed_id)
|
||||
]
|
||||
assert failed_writes
|
||||
assert failed_writes[-1][0]["status"] == IngestTaskStatus.FAILED
|
||||
assert failed_writes[-1][1] == 604800
|
||||
@@ -275,6 +300,7 @@ async def test_dedup_skipped_when_redis_unavailable() -> None:
|
||||
|
||||
async def test_dedup_lookup_failure_falls_back_to_normal_pipeline() -> None:
|
||||
"""Redis get_json 抛错：去重查询降级为未命中，走原流水线"""
|
||||
|
||||
class _ExplodingRedis(FakeRedis):
|
||||
async def get_json(self, key: str) -> dict[str, Any] | None:
|
||||
raise RuntimeError("redis down")
|
||||
|
||||
Reference in New Issue
Block a user