{
  "version": "0.2",
  "semantic": {
    "language": "zh-CN",
    "title": "Dify · 文档如何成为答案",
    "subtitle": "1.17.0 · 三层协作 · 知识库入库与基础 Chat 问答",
    "summary": "蓝色为文档入库，绿色为在线问答，灰色为共享接口。箭头表示所选调用或任务交付，不代表耗时、所有依赖或唯一执行顺序。",
    "intent": {
      "diagram_kind": "architecture",
      "primary_question": "索引和在线问答经过哪些服务，分别读写什么？",
      "audience": [
        "首次了解 Dify 架构的开发者"
      ],
      "scope": "Dify 1.17.0 自托管；本地文本上传；high_quality paragraph 索引；基础 AppMode.CHAT；已配置且选中的内部知识库、semantic_search；正常成功路径。",
      "exclusions": [
        "不覆盖整个 Dify：Workflow、Chatflow、Agent、知识流水线、多模态、重排、错误与重试不在此图。",
        "逻辑三层不是三个部署服务。Chat 与 RAG 是 API 内部代码；不是独立微服务，也不代表其他模式没有异步队列。",
        "模型调用追到 Plugin Daemon 的 HTTP 接口；其插件运行机制及下游模型厂商未在本仓库核验，故不绘制下游连线。",
        "SQL / Redis 的其他消费者、Web 服务端托管、鉴权、回调细节和缓存命中分支在概览中省略。",
        "没有运行 Dify 或调用真实模型；独立语义复核待完成。"
      ]
    },
    "entities": [
      {
        "id": "web",
        "label": "Web 控制台",
        "description": "浏览器侧\n上传 · 创建文档",
        "kind": "browser",
        "role": "process",
        "importance": "primary",
        "source_refs": [
          "src-web",
          "src-file"
        ]
      },
      {
        "id": "nginx",
        "label": "Nginx",
        "description": "反向代理\n按 URL 转发",
        "kind": "gateway",
        "role": "process",
        "importance": "primary",
        "source_refs": [
          "src-nginx"
        ]
      },
      {
        "id": "caller",
        "label": "API 调用方",
        "description": "问答请求\n流式响应",
        "kind": "client",
        "role": "process",
        "importance": "primary",
        "source_refs": [
          "src-chat-api",
          "src-nginx"
        ]
      },
      {
        "id": "api",
        "label": "API 服务",
        "description": "Chat · RAG\n进程内线程",
        "kind": "server",
        "role": "process",
        "importance": "primary",
        "source_refs": [
          "src-deploy",
          "src-dispatch",
          "src-thread",
          "src-chat",
          "src-file"
        ]
      },
      {
        "id": "worker",
        "label": "索引 Worker",
        "description": "Celery 进程\n提取 · 分段",
        "kind": "worker",
        "role": "process",
        "importance": "primary",
        "source_refs": [
          "src-deploy",
          "src-task",
          "src-index"
        ]
      },
      {
        "id": "plugin",
        "label": "插件服务",
        "description": "Plugin Daemon\n模型调用边界",
        "kind": "plugin",
        "role": "process",
        "importance": "primary",
        "source_refs": [
          "src-plugin-deploy",
          "src-plugin-llm",
          "src-plugin-embedding",
          "src-plugin-base"
        ]
      },
      {
        "id": "files",
        "label": "文件存储",
        "description": "原始文件\n默认本地存储",
        "kind": "storage",
        "role": "process",
        "importance": "primary",
        "source_refs": [
          "src-file",
          "src-extract",
          "src-deploy",
          "src-stores"
        ]
      },
      {
        "id": "redis",
        "label": "Redis",
        "description": "任务 Broker\n查询向量缓存",
        "kind": "queue",
        "role": "process",
        "importance": "primary",
        "source_refs": [
          "src-broker",
          "src-broker-default",
          "src-query-embedding"
        ]
      },
      {
        "id": "sql",
        "label": "SQL 数据库",
        "description": "元数据 · 分段\n此例 PostgreSQL",
        "kind": "database",
        "role": "process",
        "importance": "primary",
        "source_refs": [
          "src-file",
          "src-index",
          "src-postgres"
        ]
      },
      {
        "id": "vector",
        "label": "向量库",
        "description": "文档向量索引\n此例 Weaviate",
        "kind": "database",
        "role": "process",
        "importance": "primary",
        "source_refs": [
          "src-vector-create",
          "src-vector-query",
          "src-stores"
        ]
      }
    ],
    "groups": [
      {
        "id": "access",
        "label": "01  接入层 · 请求从哪里进入",
        "kind": "logical-layer",
        "members": [
          "web",
          "nginx",
          "caller"
        ]
      },
      {
        "id": "services",
        "label": "02  执行层 · 谁处理请求与索引",
        "kind": "logical-layer",
        "members": [
          "api",
          "worker",
          "plugin"
        ]
      },
      {
        "id": "data",
        "label": "03  数据层 · 内容、任务与索引",
        "kind": "logical-layer",
        "members": [
          "files",
          "redis",
          "sql",
          "vector"
        ]
      }
    ],
    "relations": [
      {
        "id": "web-nginx",
        "from": "web",
        "to": "nginx",
        "label": "上传 / 建文档",
        "kind": "http",
        "direction": "forward",
        "condition": "Console request succeeds",
        "importance": "primary",
        "source_refs": [
          "src-web",
          "src-file",
          "src-nginx"
        ]
      },
      {
        "id": "caller-nginx",
        "from": "caller",
        "to": "nginx",
        "label": "问答 / SSE",
        "kind": "http",
        "direction": "bidirectional",
        "condition": "Basic CHAT; streaming response mode",
        "importance": "primary",
        "source_refs": [
          "src-nginx",
          "src-chat-api"
        ]
      },
      {
        "id": "nginx-api",
        "from": "nginx",
        "to": "api",
        "label": "API 请求 / 响应",
        "kind": "http",
        "direction": "bidirectional",
        "condition": "Console API or /v1 route",
        "importance": "primary",
        "source_refs": [
          "src-nginx",
          "src-chat-api"
        ]
      },
      {
        "id": "api-files",
        "from": "api",
        "to": "files",
        "label": "保存文件",
        "kind": "storage-access",
        "direction": "forward",
        "condition": "Local uploaded text file; validation succeeds",
        "importance": "primary",
        "source_refs": [
          "src-file"
        ]
      },
      {
        "id": "api-sql",
        "from": "api",
        "to": "sql",
        "label": "保存元数据",
        "kind": "database-access",
        "direction": "forward",
        "condition": "Upload and document creation succeed",
        "importance": "primary",
        "source_refs": [
          "src-file",
          "src-document"
        ]
      },
      {
        "id": "api-redis",
        "from": "api",
        "to": "redis",
        "label": "任务 / 查询缓存",
        "kind": "dependency",
        "direction": "forward",
        "condition": "Index dispatch after document commit; query embedding cache access on semantic retrieval",
        "importance": "primary",
        "source_refs": [
          "src-document",
          "src-task-proxy",
          "src-task-policy",
          "src-broker",
          "src-broker-default",
          "src-query-embedding"
        ]
      },
      {
        "id": "redis-worker",
        "from": "redis",
        "to": "worker",
        "label": "领取索引任务",
        "kind": "task-delivery",
        "direction": "forward",
        "condition": "Worker consumes Celery broker; self-hosted priority_dataset queue",
        "importance": "primary",
        "source_refs": [
          "src-broker",
          "src-task-policy",
          "src-task",
          "src-task-run"
        ]
      },
      {
        "id": "worker-files",
        "from": "worker",
        "to": "files",
        "label": "读取原文件",
        "kind": "storage-access",
        "direction": "forward",
        "condition": "FILE source in extraction",
        "importance": "primary",
        "source_refs": [
          "src-task-run",
          "src-index",
          "src-extract"
        ]
      },
      {
        "id": "worker-sql",
        "from": "worker",
        "to": "sql",
        "label": "保存分段",
        "kind": "database-access",
        "direction": "forward",
        "condition": "Successful paragraph indexing",
        "importance": "primary",
        "source_refs": [
          "src-index",
          "src-doc-embedding"
        ]
      },
      {
        "id": "worker-vector",
        "from": "worker",
        "to": "vector",
        "label": "写入索引",
        "kind": "vector-access",
        "direction": "forward",
        "condition": "high_quality paragraph indexing; embedding succeeds",
        "importance": "primary",
        "source_refs": [
          "src-index",
          "src-paragraph",
          "src-vector-create"
        ]
      },
      {
        "id": "api-vector",
        "from": "api",
        "to": "vector",
        "label": "语义检索",
        "kind": "vector-access",
        "direction": "forward",
        "condition": "Dataset configured and selected; internal high_quality semantic retrieval; filters permit results",
        "importance": "primary",
        "source_refs": [
          "src-chat",
          "src-retrieval",
          "src-search",
          "src-vector-query"
        ]
      },
      {
        "id": "worker-plugin",
        "from": "worker",
        "to": "plugin",
        "label": "文档向量化",
        "kind": "model-invocation",
        "direction": "forward",
        "condition": "Document embedding cache miss; model configured",
        "importance": "primary",
        "source_refs": [
          "src-vector-create",
          "src-doc-embedding",
          "src-model",
          "src-plugin-embedding",
          "src-plugin-base"
        ]
      },
      {
        "id": "api-plugin",
        "from": "api",
        "to": "plugin",
        "label": "Embedding / LLM",
        "kind": "model-invocation",
        "direction": "forward",
        "condition": "Query embedding cache miss and/or LLM invocation; validated normal CHAT path",
        "importance": "primary",
        "source_refs": [
          "src-chat",
          "src-query-embedding",
          "src-model",
          "src-plugin-runtime",
          "src-plugin-llm",
          "src-plugin-embedding",
          "src-plugin-base"
        ]
      }
    ],
    "flows": [
      {
        "id": "index-path",
        "label": "索引投递与写入",
        "importance": "primary",
        "repeat": "event-driven",
        "relation_ids": [
          "web-nginx",
          "nginx-api",
          "api-redis",
          "redis-worker",
          "worker-vector"
        ]
      },
      {
        "id": "retrieval-path",
        "label": "问答中的检索调用",
        "importance": "primary",
        "repeat": "event-driven",
        "relation_ids": [
          "caller-nginx",
          "nginx-api",
          "api-vector"
        ]
      }
    ],
    "sources": [
      {
        "id": "src-web",
        "type": "repository",
        "title": "Web creates dataset documents",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "web/service/datasets.ts",
          "line": 153,
          "end_line": 170
        }
      },
      {
        "id": "src-nginx",
        "type": "repository",
        "title": "Nginx API routes",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "docker/nginx/conf.d/default.conf.template",
          "line": 8,
          "end_line": 45
        }
      },
      {
        "id": "src-deploy",
        "type": "repository",
        "title": "API and Worker are separate services",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "docker/docker-compose.yaml",
          "line": 227,
          "end_line": 352
        }
      },
      {
        "id": "src-plugin-deploy",
        "type": "repository",
        "title": "Plugin daemon service",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "docker/docker-compose.yaml",
          "line": 573,
          "end_line": 602
        }
      },
      {
        "id": "src-file",
        "type": "repository",
        "title": "File bytes and SQL metadata are separate",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/services/file_service.py",
          "line": 37,
          "end_line": 125
        }
      },
      {
        "id": "src-document",
        "type": "repository",
        "title": "Commit documents before submitting indexing",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/services/dataset_service.py",
          "line": 2482,
          "end_line": 2499
        }
      },
      {
        "id": "src-broker",
        "type": "repository",
        "title": "Celery broker configuration",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/extensions/ext_celery.py",
          "line": 113,
          "end_line": 132
        }
      },
      {
        "id": "src-broker-default",
        "type": "repository",
        "title": "Redis broker default",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "docker/.env.example",
          "line": 115,
          "end_line": 124
        }
      },
      {
        "id": "src-stores",
        "type": "repository",
        "title": "Storage and vector store defaults",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "docker/.env.example",
          "line": 159,
          "end_line": 164
        }
      },
      {
        "id": "src-postgres",
        "type": "repository",
        "title": "PostgreSQL deployment option",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "docker/docker-compose.yaml",
          "line": 425,
          "end_line": 437
        }
      },
      {
        "id": "src-task-proxy",
        "type": "repository",
        "title": "Direct Celery dispatch",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/services/document_indexing_proxy/batch_indexing_base.py",
          "line": 37,
          "end_line": 47
        }
      },
      {
        "id": "src-task-policy",
        "type": "repository",
        "title": "Self-hosted dispatch policy",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/services/document_indexing_proxy/base.py",
          "line": 79,
          "end_line": 111
        }
      },
      {
        "id": "src-task",
        "type": "repository",
        "title": "Indexing task entry points",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/tasks/document_indexing_task.py",
          "line": 200,
          "end_line": 267
        }
      },
      {
        "id": "src-task-run",
        "type": "repository",
        "title": "Task invokes IndexingRunner",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/tasks/document_indexing_task.py",
          "line": 49,
          "end_line": 125
        }
      },
      {
        "id": "src-index",
        "type": "repository",
        "title": "Extract, transform, persist segments, load index",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/indexing_runner.py",
          "line": 81,
          "end_line": 159
        }
      },
      {
        "id": "src-extract",
        "type": "repository",
        "title": "Download uploaded file for extraction",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/rag/extractor/extract_processor.py",
          "line": 115,
          "end_line": 137
        }
      },
      {
        "id": "src-paragraph",
        "type": "repository",
        "title": "High-quality paragraph vector indexing",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/rag/index_processor/processor/paragraph_index_processor.py",
          "line": 134,
          "end_line": 149
        }
      },
      {
        "id": "src-vector-create",
        "type": "repository",
        "title": "Embed documents then write vectors",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/rag/datasource/vdb/vector_factory.py",
          "line": 168,
          "end_line": 189
        }
      },
      {
        "id": "src-vector-query",
        "type": "repository",
        "title": "Embed query then search vectors",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/rag/datasource/vdb/vector_factory.py",
          "line": 252,
          "end_line": 259
        }
      },
      {
        "id": "src-doc-embedding",
        "type": "repository",
        "title": "Document embedding SQL cache",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/rag/embedding/cached_embedding.py",
          "line": 29,
          "end_line": 103
        }
      },
      {
        "id": "src-query-embedding",
        "type": "repository",
        "title": "Query embedding Redis cache",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/rag/embedding/cached_embedding.py",
          "line": 194,
          "end_line": 241
        }
      },
      {
        "id": "src-chat-api",
        "type": "repository",
        "title": "Generate chat and return response stream",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/controllers/service_api/app/completion.py",
          "line": 388,
          "end_line": 419
        }
      },
      {
        "id": "src-dispatch",
        "type": "repository",
        "title": "Basic CHAT dispatch",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/services/app_generate_service.py",
          "line": 218,
          "end_line": 231
        }
      },
      {
        "id": "src-thread",
        "type": "repository",
        "title": "API-local thread and ChatAppRunner",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/app/apps/chat/app_generator.py",
          "line": 208,
          "end_line": 272
        }
      },
      {
        "id": "src-chat",
        "type": "repository",
        "title": "Retrieve context then generate answer",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/app/apps/chat/app_runner.py",
          "line": 160,
          "end_line": 256
        }
      },
      {
        "id": "src-retrieval",
        "type": "repository",
        "title": "Selected internal dataset retrieval",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/rag/retrieval/dataset_retrieval.py",
          "line": 660,
          "end_line": 754
        }
      },
      {
        "id": "src-search",
        "type": "repository",
        "title": "Semantic vector retrieval",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/rag/datasource/retrieval_service.py",
          "line": 303,
          "end_line": 347
        }
      },
      {
        "id": "src-model",
        "type": "repository",
        "title": "LLM and embedding model invocation",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/model_manager.py",
          "line": 181,
          "end_line": 266
        }
      },
      {
        "id": "src-plugin-runtime",
        "type": "repository",
        "title": "Model runtime delegates to plugin client",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/plugin/impl/model_runtime.py",
          "line": 309,
          "end_line": 351
        }
      },
      {
        "id": "src-plugin-llm",
        "type": "repository",
        "title": "Plugin daemon LLM HTTP contract",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/plugin/impl/model.py",
          "line": 170,
          "end_line": 217
        }
      },
      {
        "id": "src-plugin-embedding",
        "type": "repository",
        "title": "Plugin daemon embedding HTTP contract",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/plugin/impl/model.py",
          "line": 360,
          "end_line": 399
        }
      },
      {
        "id": "src-plugin-base",
        "type": "repository",
        "title": "HTTP client points at plugin daemon",
        "repository": {
          "url": "https://github.com/langgenius/dify",
          "revision": "09a855dcef24c0edc7431c46c0cfaa494481daf5",
          "path": "api/core/plugin/impl/base.py",
          "line": 98,
          "end_line": 165
        }
      }
    ]
  },
  "presentation": {
    "icon_system": "auto",
    "style": "deep-tech",
    "layout": "layered",
    "motion": "showcase-v1"
  },
  "presentation_sources": {
    "style": "explicit",
    "layout": "model"
  },
  "reader": {
    "enabled": true,
    "views": [
      {
        "id": "ingest",
        "label": "文档入库",
        "relation_ids": [
          "web-nginx",
          "nginx-api",
          "api-files",
          "api-sql",
          "api-redis",
          "redis-worker",
          "worker-files",
          "worker-sql",
          "worker-vector",
          "worker-plugin"
        ]
      },
      {
        "id": "query",
        "label": "在线问答",
        "relation_ids": [
          "caller-nginx",
          "nginx-api",
          "api-redis",
          "api-vector",
          "api-plugin"
        ]
      }
    ]
  }
}