#知识库文档处理解析

flowchart LR
    A[文档上传] --> B[Ingest 接收]
    B --> C[Parse 解析]
    C --> D{异步解析?}
    D -->|是| E[提交异步任务]
    D -->|否| F[Chunk 切片]
    E --> G[等待回调]
    G --> F
    F --> H[Summary 摘要]
    F --> I[Tag 标签]
    H --> J[Embed 向量化]
    I --> J
    J --> K[完成]

flowchart LR
    A[文档] --> B[遍历解析器]
    B --> C{supports?}
    C -->|否| B
    C -->|是| D[加入候选列表]
    D --> E[按优先级排序]
    E --> F[选择最高优先级]
    F --> G[执行解析]

flowchart LR
    A[请求入口] --> B{文档类型?}
    B -->|文件上传| C[ingestDocument]
    B -->|文本录入| D[ingestText]
    B -->|Excel QA| E[ingestExcelQa]
    B -->|手动问答| F[ingestManualQa]
    C --> G[加载文档实体]
    D --> G
    E --> G
    F --> G
    G --> H[加载知识库配置]
    H --> I[初始化上下文]

flowchart LR
    A[选择解析器] --> B[执行解析]
    B --> C{解析状态?}
    C -->|SUCCESS| D[数据清洗]
    C -->|ASYNC| E[保存任务ID]
    C -->|FAIL| F[记录错误]
    D --> G[设置 rawContent]
    E --> H[提交异步任务]
    H --> I[标记解析中]
    F --> J[标记失败]
    G --> K[进入 Chunk]

flowchart LR
    A[Excel 文件] --> B[POI 读取]
    B --> C[遍历行数据]
    C --> D[提取 Q/A 列]
    D --> E[构建 QaItem]
    E --> F[设置 qaList]
    F --> G[创建切片]

flowchart LR
    A[rawContent] --> B{切片策略?}
    B -->|PARAGRAPH| C[按段落分割]
    B -->|CHARACTER| D[按字符分割]
    B -->|SENTENCE| E[按句子分割]
    B -->|默认| F[递归分割]
    C --> G[生成切片列表]
    D --> G
    E --> G
    F --> G
    G --> H[批量保存]
    H --> I[设置 slices]

flowchart LR
    A[切片列表] --> B[选择向量化策略]
    B --> C{知识库模式?}
    C -->|Q2Q 客服| D[纯问题向量化]
    C -->|通用| E[文档名+内容+摘要]
    D --> F[构建 TextSegment]
    E --> F
    F --> G[调用 EmbeddingModel]
    G --> H[存入向量库]
    H --> I[更新切片状态]

flowchart LR
    A[处理结果] --> B{状态?}
    B -->|异步解析| C[markParsing]
    B -->|处理完成| D[markCompleted]
    B -->|处理失败| E[markFailed]
    C --> F[ASYNC_PARSING]
    D --> G[SLICED]
    E --> H[FAILED]

flowchart LR
    A[ingestDocument] --> B[selectParser]
    B --> C[parseDocument]
    C --> D{isAsyncParser?}
    D -->|是| E[submitAsyncTask]
    E --> F[markParsing]
    D -->|否| G{parseSuccess?}
    G -->|是| H[chunkDocument]
    G -->|否| I[markFailed]
    H --> J[summaryDocument]
    H --> K[tagDocument]
    J --> L[embedSlices]
    K --> L
    L --> M[markCompleted]

flowchart LR
    A[ingestExcelQa] --> B[parseExcelQa]
    B --> C[createQaSlices]
    C --> D[embedSlices]
    D --> E[markCompleted]

flowchart LR
    A[EmbeddingStoreFactory] --> B{向量库类型?}
    B -->|milvus| C[MilvusEmbeddingStoreFactory]
    B -->|qdrant| D[QdrantEmbeddingStoreFactory]
    B -->|chroma| E[ChromaEmbeddingStoreFactory]
    B -->|pgvector| F[PgVectorEmbeddingStoreFactory]
    B -->|neo4j| G[Neo4jEmbeddingStoreFactory]
    C --> H[DefaultEmbeddingStrategy]
    D --> H
    E --> H
    F --> H
    G --> I[Neo4jEmbeddingStrategy]