feat(kb): 接入 Dify 知识库，移除 FastGPT；AI 聊天产出+用户文档自动入库量化

- 新增 app/dify_kb.py Dify v1 客户端(连通测试/检索/文本入库)
- kb_router 仅保留 Dify 外部检索分支
- knowledge_admin 去掉 FastGPT/RAGFlow，新增 dify 批量入库端点
- chat.py 注入 Dify RAG 上下文 + 助手回答 fire-and-forget 入库
- 前端 Knowledge.tsx 改为 Dify 单服务 UI(可通讯/可用验证+入库按钮)
- DB: external_kb_services 删除 fastgpt，qihuai 设为 https+connected

backend/python/app/api/admin/knowledge_admin.py

@@ -32,7 +32,7 @@ logger = logging.getLogger("aiflow")
 
 class ExternalServiceConfig(BaseModel):
     name: str = Field(min_length=1, max_length=200)
-    service_type: str = Field(..., description="fastgpt / dify / ragflow")
+    service_type: str = Field(..., description="dify")
     api_url: str = Field(min_length=1, max_length=500)
     api_key: Optional[str] = None
     config: Optional[Dict[str, Any]] = None
@@ -739,48 +739,27 @@ def _get_external_services_table(db: Session):
 
 
 def _check_service_connection(service_type: str, api_url: str, api_key: str) -> dict:
-    """测试外部知识库服务连接状态"""
+    """测试外部知识库服务连接状态（当前仅支持 Dify）"""
     try:
+        if service_type == "dify":
+            from ...dify_kb import dify_test
+            return dify_test(api_url, api_key)
+        else:
+            # 未知类型（如历史遗留的 fastgpt/ragflow），按通用方式探测可达性
             headers = {}
             if api_key:
                 headers["Authorization"] = f"Bearer {api_key}"
-
-        if service_type == "fastgpt":
-            # FastGPT 健康检查
-            check_url = f"{api_url.rstrip('/')}/api/v1/status"
-            resp = httpx.get(check_url, headers=headers, timeout=10)
-            if resp.status_code == 200:
-                return {"connected": True, "status": "connected", "detail": "FastGPT 连接成功"}
-            else:
-                return {"connected": False, "status": "error", "detail": f"FastGPT 返回状态码 {resp.status_code}"}
-
-        elif service_type == "dify":
-            # Dify 健康检查
-            check_url = f"{api_url.rstrip('/')}/v1/datasets"
-            resp = httpx.get(check_url, headers=headers, timeout=10)
-            if resp.status_code in [200, 401]:  # 401说明服务存在但需要认证
-                return {"connected": True, "status": "connected", "detail": "Dify 连接成功"}
-            else:
-                return {"connected": False, "status": "error", "detail": f"Dify 返回状态码 {resp.status_code}"}
-
-        elif service_type == "ragflow":
-            # RAGFlow 健康检查
-            check_url = f"{api_url.rstrip('/')}/api/v1/datasets"
-            resp = httpx.get(check_url, headers=headers, timeout=10)
-            if resp.status_code in [200, 401]:
-                return {"connected": True, "status": "connected", "detail": "RAGFlow 连接成功"}
-            else:
-                return {"connected": False, "status": "error", "detail": f"RAGFlow 返回状态码 {resp.status_code}"}
-
-        else:
-            # 通用检查
             resp = httpx.get(api_url, headers=headers, timeout=10)
-            return {"connected": resp.status_code < 500, "status": "connected" if resp.status_code < 500 else "error"}
+            return {
-
+                "connected": resp.status_code < 500,
+                "usable": resp.status_code < 500,
+                "status": "connected" if resp.status_code < 500 else "error",
+                "detail": f"通用探测返回状态码 {resp.status_code}",
+            }
     except httpx.TimeoutException:
-        return {"connected": False, "status": "timeout", "detail": "连接超时"}
+        return {"connected": False, "usable": False, "status": "timeout", "detail": "连接超时"}
     except Exception as e:
-        return {"connected": False, "status": "error", "detail": str(e)}
+        return {"connected": False, "usable": False, "status": "error", "detail": str(e)}
 
 
 @router.get("/external/services", summary="获取外部知识库服务列表")
@@ -801,9 +780,7 @@ async def list_external_services(
             "total": len(services),
             "items": services,
             "supported_types": [
-                {"type": "fastgpt", "name": "FastGPT", "description": "开源 AI 知识库平台，支持多种向量模型"},
+                {"type": "dify", "name": "Dify", "description": "LLM 应用开发平台，内置知识库管理（RAG）"},
-                {"type": "dify", "name": "Dify", "description": "LLM 应用开发平台，内置知识库管理"},
-                {"type": "ragflow", "name": "RAGFlow", "description": "基于深度文档理解的开源 RAG 引擎"},
             ]
         }
     }
@@ -816,8 +793,8 @@ async def create_external_service(
     db: Session = Depends(get_db),
 ):
     """创建新的外部知识库服务连接配置"""
-    # 验证服务类型
+    # 验证服务类型（当前仅支持 Dify）
-    valid_types = ["fastgpt", "dify", "ragflow"]
+    valid_types = ["dify"]
     if req.service_type not in valid_types:
         raise HTTPException(status_code=400, detail=f"不支持的服务类型: {req.service_type}")
 
@@ -1082,163 +1059,30 @@ async def sync_knowledge_base(
         if api_key:
             headers["Authorization"] = f"Bearer {api_key}"
 
-        if service_type == "fastgpt" and req.sync_direction == "push":
+        if service_type == "dify" and req.sync_direction == "push":
-            # 向 FastGPT 推送知识库数据
+            # 向 Dify 推送知识库数据（使用 app.dify_kb 客户端，按 dataset_name 解析）
-            # 1. 在 FastGPT 创建/获取对应的 dataset
+            from ...dify_kb import dify_ingest_text, resolve_dataset_id
-            ext_kb_id = getattr(kb, "external_id", None)
+            dataset_name = service_config.get("dataset_name") if isinstance(service_config, dict) else None
-
+            ds_id = resolve_dataset_id(api_url, api_key, dataset_name)
-            if not ext_kb_id:
+            if not ds_id:
-                # 创建新 dataset
+                sync_result["detail"] = "无法解析 Dify 知识库 ID（请检查 API Key 是否绑定到指定数据集）"
-                create_resp = httpx.post(
-                    f"{api_url.rstrip('/')}/api/v1/datasets",
-                    headers=headers,
-                    json={"name": kb.name, "description": kb.description or ""},
-                    timeout=30,
-                )
-                if create_resp.status_code == 200:
-                    ext_data = create_resp.json()
-                    ext_kb_id = ext_data.get("data", {}).get("id") or ext_data.get("id")
-                    # 更新 KB 的 external_id
-                    try:
-                        setattr(kb, "external_id", ext_kb_id)
-                    except Exception:
-                        db.execute(
-                            text("UPDATE knowledge_bases SET external_id = :eid WHERE id = :kid"),
-                            {"eid": ext_kb_id, "kid": kb_id}
-                        )
-                else:
-                    sync_result["detail"] = f"FastGPT 创建 dataset 失败: {create_resp.status_code}"
                 sync_result["synced"] = False
-
+            else:
-            if ext_kb_id:
-                # 推送文档内容到 FastGPT
                 docs = db.query(KnowledgeDocument).filter(
                     KnowledgeDocument.knowledge_base_id == kb_id,
                     KnowledgeDocument.status == "completed"
                 ).all()
-
                 synced_count = 0
                 for doc in docs:
                     content = doc.content or ""
                     if not content:
                         continue
-
+                    if dify_ingest_text(api_url, api_key, content[:50000], name=doc.filename, dataset_name=dataset_name):
-                    # 分段推送
-                    push_resp = httpx.post(
-                        f"{api_url.rstrip('/')}/api/v1/datasets/{ext_kb_id}/documents",
-                        headers=headers,
-                        json={
-                            "name": doc.filename,
-                            "content": content[:50000],  # 限制单次推送大小
-                        },
-                        timeout=30,
-                    )
-                    if push_resp.status_code in [200, 201]:
                         synced_count += 1
-
-                sync_result["synced"] = True
-                sync_result["detail"] = f"已推送 {synced_count}/{len(docs)} 文档到 FastGPT"
-                sync_result["items_synced"] = synced_count
-
-        elif service_type == "dify" and req.sync_direction == "push":
-            # 向 Dify 推送知识库数据
-            ext_kb_id = getattr(kb, "external_id", None)
-
-            if not ext_kb_id:
-                create_resp = httpx.post(
-                    f"{api_url.rstrip('/')}/v1/datasets",
-                    headers=headers,
-                    json={"name": kb.name},
-                    timeout=30,
-                )
-                if create_resp.status_code in [200, 201]:
-                    ext_data = create_resp.json()
-                    ext_kb_id = ext_data.get("id")
-                    try:
-                        setattr(kb, "external_id", ext_kb_id)
-                    except Exception:
-                        db.execute(
-                            text("UPDATE knowledge_bases SET external_id = :eid WHERE id = :kid"),
-                            {"eid": ext_kb_id, "kid": kb_id}
-                        )
-
-            if ext_kb_id:
-                docs = db.query(KnowledgeDocument).filter(
-                    KnowledgeDocument.knowledge_base_id == kb_id,
-                    KnowledgeDocument.status == "completed"
-                ).all()
-
-                synced_count = 0
-                for doc in docs:
-                    content = doc.content or ""
-                    if not content:
-                        continue
-
-                    # Dify 文档创建
-                    doc_resp = httpx.post(
-                        f"{api_url.rstrip('/')}/v1/datasets/{ext_kb_id}/document/create_by_text",
-                        headers=headers,
-                        json={
-                            "name": doc.filename,
-                            "text": content[:50000],
-                            "indexing_technique": "high_quality",
-                        },
-                        timeout=30,
-                    )
-                    if doc_resp.status_code in [200, 201]:
-                        synced_count += 1
-
                 sync_result["synced"] = True
                 sync_result["detail"] = f"已推送 {synced_count}/{len(docs)} 文档到 Dify"
                 sync_result["items_synced"] = synced_count
 
-        elif service_type == "ragflow" and req.sync_direction == "push":
-            # 向 RAGFlow 推送知识库数据
-            ext_kb_id = getattr(kb, "external_id", None)
-
-            if not ext_kb_id:
-                create_resp = httpx.post(
-                    f"{api_url.rstrip('/')}/api/v1/datasets",
-                    headers=headers,
-                    json={"name": kb.name},
-                    timeout=30,
-                )
-                if create_resp.status_code in [200, 201]:
-                    ext_data = create_resp.json()
-                    ext_kb_id = ext_data.get("id")
-                    try:
-                        setattr(kb, "external_id", ext_kb_id)
-                    except Exception:
-                        db.execute(
-                            text("UPDATE knowledge_bases SET external_id = :eid WHERE id = :kid"),
-                            {"eid": ext_kb_id, "kid": kb_id}
-                        )
-
-            if ext_kb_id:
-                docs = db.query(KnowledgeDocument).filter(
-                    KnowledgeDocument.knowledge_base_id == kb_id,
-                    KnowledgeDocument.status == "completed"
-                ).all()
-
-                synced_count = 0
-                for doc in docs:
-                    content = doc.content or ""
-                    if not content:
-                        continue
-
-                    doc_resp = httpx.post(
-                        f"{api_url.rstrip('/')}/api/v1/datasets/{ext_kb_id}/documents",
-                        headers=headers,
-                        json={"name": doc.filename, "content": content[:50000]},
-                        timeout=30,
-                    )
-                    if doc_resp.status_code in [200, 201]:
-                        synced_count += 1
-
-                sync_result["synced"] = True
-                sync_result["detail"] = f"已推送 {synced_count}/{len(docs)} 文档到 RAGFlow"
-                sync_result["items_synced"] = synced_count
-
         elif req.sync_direction == "pull":
             # 从外部服务拉取知识库数据
             sync_result["detail"] = f"从 {service_type} 拉取数据（暂未实现完整拉取逻辑）"
@@ -1288,6 +1132,106 @@ async def sync_knowledge_base(
     return {"code": 0, "data": sync_result}
 
 
+# ============================================================
+# 平台内容批量入库到 Dify（AI聊天产出 + 用户上传文档）
+# ============================================================
+
+@router.post("/external/dify/ingest", summary="将平台内容批量入库到 Dify 知识库")
+async def ingest_platform_to_dify(
+    admin_user=Depends(get_current_admin),
+    db: Session = Depends(get_db),
+):
+    """把平台中「可量化的内容」推送到 Dify 知识库（qihuai）:
+       - AI 聊天的助手回答（文本、非工具/非媒体、长度 > 120）
+       - 用户上传的文档资产（assets, category='file' 且含文本内容）
+    返回各来源的入库数量。
+    """
+    from ...dify_kb import dify_ingest_text, resolve_dataset_id
+
+    # 读取已启用的 Dify 服务
+    try:
+        row = db.execute(
+            text("SELECT api_url, api_key, config FROM external_kb_services "
+                 "WHERE service_type='dify' AND is_enabled=1 LIMIT 1")
+        ).fetchone()
+    except Exception:
+        row = None
+    if not row:
+        raise HTTPException(status_code=400, detail="未配置已启用的 Dify 知识库服务，请先添加 Dify 服务并测试连接")
+    api_url, api_key, config_raw = row
+    dataset_name = None
+    if config_raw:
+        try:
+            cfg = config_raw if isinstance(config_raw, dict) else json.loads(config_raw)
+            dataset_name = cfg.get("dataset_name")
+        except Exception:
+            pass
+    ds_id = resolve_dataset_id(api_url, api_key, dataset_name)
+    if not ds_id:
+        raise HTTPException(status_code=400, detail="无法解析 Dify 知识库 ID（请检查 API Key 与 dataset_name）")
+
+    stats = {"chat_answers": 0, "user_docs": 0, "skipped": 0, "failed": 0}
+
+    # 1. AI 聊天助手回答
+    try:
+        from ...models import Message
+        answers = db.query(Message).filter(
+            Message.role == "assistant"
+        ).order_by(Message.created_at.asc()).limit(2000).all()
+        for m in answers:
+            content = getattr(m, "content", None) or ""
+            content = content.strip()
+            # 跳过媒体/工具类消息与过短内容
+            if len(content) < 120:
+                stats["skipped"] += 1
+                continue
+            if content.startswith("[") and ("MEDIA" in content or "IMAGE" in content or "VIDEO" in content):
+                stats["skipped"] += 1
+                continue
+            conv_id = getattr(m, "conversation_id", None) or "unknown"
+            if dify_ingest_text(api_url, api_key, content, name=f"chat_{conv_id}_{m.id}", dataset_name=dataset_name):
+                stats["chat_answers"] += 1
+            else:
+                stats["failed"] += 1
+    except Exception as e:
+        logger.error(f"ingest chat answers failed: {e}")
+
+    # 2. 用户上传的文档资产
+    try:
+        from ...models import Asset
+        assets = db.query(Asset).filter(
+            getattr(Asset, "category", "file") == "file"
+        ).order_by(getattr(Asset, "created_at", text("id")).asc()).limit(2000).all()
+        for a in assets:
+            content = getattr(a, "content", None) or ""
+            content = content.strip()
+            if len(content) < 50:
+                stats["skipped"] += 1
+                continue
+            fname = getattr(a, "filename", None) or getattr(a, "name", None) or "asset"
+            if dify_ingest_text(api_url, api_key, content[:50000], name=f"asset_{a.id}_{fname}", dataset_name=dataset_name):
+                stats["user_docs"] += 1
+            else:
+                stats["failed"] += 1
+    except Exception as e:
+        logger.error(f"ingest assets failed: {e}")
+
+    LogService.log_audit(
+        admin_id=admin_user.id,
+        admin_name=admin_user.username,
+        action="ingest_platform_to_dify",
+        target_type="external_kb_service",
+        target_id=ds_id,
+        detail=f"平台内容入库 Dify: 聊天回答 {stats['chat_answers']} 条, 用户文档 {stats['user_docs']} 条, 跳过 {stats['skipped']}, 失败 {stats['failed']}",
+    )
+
+    return {
+        "code": 0,
+        "message": f"入库完成：聊天回答 {stats['chat_answers']} 条，用户文档 {stats['user_docs']} 条",
+        "data": stats,
+    }
+
+
 # ============================================================
 # 批量操作
 # ============================================================

backend/python/app/api/chat.py

@@ -490,14 +490,13 @@ def _retrieve_kb_context(db: Session, kb_ids: list, query: str, max_chars: int =
     2. 构建文档列表(content+title+metadata)
     3. BM25检索最相关的段落
     4. 拼接为上下文注入system prompt
+    同时叠加 Dify 外部知识库(RAG)检索结果。
     """
-    if not kb_ids:
+    local_context = ""
-        return ""
 
+    if kb_ids:
         kbs = db.query(KnowledgeBase).filter(KnowledgeBase.id.in_(kb_ids)).all()
-    if not kbs:
+        if kbs:
-        return ""
-
             # 收集所有文档段落
             # Stage C.1: 优先使用 KnowledgeDocumentChunk 表的预分段数据
             all_segments = []
@@ -554,20 +553,59 @@ def _retrieve_kb_context(db: Session, kb_ids: list, query: str, max_chars: int =
             if not all_segments:
                 # 退化: 使用原始朴素拼接
                 kb_names = ", ".join([k.name for k in kbs])
-        return f"参考知识库: {kb_names}"
+                local_context = f"参考知识库: {kb_names}"
-
+            else:
                 # BM25检索最相关的段落
                 top_segments = _bm25_search(query, all_segments, top_k=8)
-
-    # 拼接上下文
                 context_parts = [seg["content"] for seg in top_segments]
-    context = "\n".join(context_parts)
+                local_context = "\n".join(context_parts)
+                if len(local_context) > max_chars:
+                    local_context = local_context[:max_chars] + "...(内容截断)"
 
-    # 限制总长度
+    # 叠加 Dify 外部知识库(RAG)检索
-    if len(context) > max_chars:
+    dify_context = ""
-        context = context[:max_chars] + "...(内容截断)"
+    try:
+        from app.dify_kb import get_dify_kb_context
+        dify_context = get_dify_kb_context(db, query, top_k=5)
+    except Exception as e:
+        logger.warning(f"Dify KB context retrieval failed: {e}")
 
-    return context
+    parts = []
+    if local_context:
+        parts.append(local_context)
+    if dify_context:
+        parts.append("【Dify 知识库检索结果】\n" + dify_context)
+    return "\n\n".join(parts)
+
+
+async def _ingest_answer_to_dify(conv_id: str, msg_id: str, content: str):
+    """后台任务：将助手回答入库到 Dify 知识库（量化后可供 RAG 检索复用）。"""
+    try:
+        from app.dify_kb import aingest_text
+        from app.database import SessionLocal
+        from sqlalchemy import text as _t
+        db = SessionLocal()
+        try:
+            row = db.execute(_t(
+                "SELECT api_url, api_key, config FROM external_kb_services "
+                "WHERE service_type='dify' AND is_enabled=1 LIMIT 1"
+            )).fetchone()
+        finally:
+            db.close()
+        if not row:
+            return
+        api_url, api_key, config_raw = row
+        dataset_name = None
+        if config_raw:
+            try:
+                import json
+                cfg = config_raw if isinstance(config_raw, dict) else json.loads(config_raw)
+                dataset_name = cfg.get("dataset_name")
+            except Exception:
+                pass
+        await aingest_text(api_url, api_key, content, name=f"chat_{conv_id}_{msg_id}", dataset_name=dataset_name)
+    except Exception as e:
+        logger.warning(f"后台入库 Dify 失败: {e}")
 
 
 # ============================================================
@@ -1203,6 +1241,20 @@ async def send_message(
     )
     db.add(ai_msg)
 
+    # 后台异步：将实质性文本回答入库到 Dify 知识库（量化后可供 RAG 复用）
+    _ingest_text = (ai_content or "").strip()
+    if len(_ingest_text) > 120 and not (
+        _ingest_text.startswith("[MEDIA")
+        or _ingest_text.startswith("data:")
+        or _ingest_text.startswith("http")
+        or "<img" in _ingest_text
+        or "![]( " in _ingest_text
+    ):
+        try:
+            asyncio.create_task(_ingest_answer_to_dify(conv_id, ai_msg.id, _ingest_text))
+        except Exception:
+            pass
+
     # 更新used_skills
     used_skills = []
     if skills:

backend/python/app/dify_kb.py

@@ -0,0 +1,248 @@
+"""
+Dify 知识库客户端
+================
+封装与 Dify 知识库(v1 API)的交互：连通性测试、检索(RAG)、文本入库。
+设计用于替换原 FastGPT 外部知识库对接。
+
+已验证的接入点（2026-07-14 实测）:
+- Base:  https://dify.qihuyun.com/v1
+         （直接 HTTP 访问会被 nginx 301 跳转到 HTTPS，必须 HTTPS 直连，
+          且内网自签证书需 verify=False）
+- Auth:  Bearer {api_key}
+- GET  /v1/datasets                               列出知识库(可据此解析 dataset id)
+- POST /v1/datasets/{id}/document/create-by-text  文本入库
+- POST /v1/datasets/{id}/document/create-by-file  文件入库(multipart)
+- POST /v1/datasets/{id}/retrieve                 检索(RAG)
+  retrieve 必需完整的 retrieval_model 字典，否则返回 400。
+"""
+import logging
+import threading
+import time
+import httpx
+
+logger = logging.getLogger("aiflow.dify_kb")
+
+# dataset id 缓存: key -> (dataset_id, ts)
+_dataset_cache: dict = {}
+_dataset_cache_lock = threading.Lock()
+_DATASET_CACHE_TTL = 3600
+
+
+def _normalize_base(api_url: str) -> str:
+    """统一 base url -> https 形式，且不带 /v1 后缀。"""
+    u = (api_url or "").strip()
+    if not u:
+        u = "https://dify.qihuyun.com/v1"
+    u = u.rstrip("/")
+    if u.endswith("/v1"):
+        u = u[: -len("/v1")]
+    if u.startswith("http://"):
+        u = "https://" + u[len("http://"):]
+    return u
+
+
+def dify_test(api_url: str, api_key: str) -> dict:
+    """验证 Dify 知识库「可通讯 + 可用」。
+
+    返回:
+        {
+          "connected": bool,
+          "usable": bool,
+          "status": "connected" | "auth_error" | "timeout" | "error",
+          "detail": str,
+          "datasets": [{"id":..., "name":...}, ...]
+        }
+    """
+    base = _normalize_base(api_url)
+    headers = {"Authorization": f"Bearer {api_key}"}
+    try:
+        with httpx.Client(base_url=base, timeout=15, verify=False) as client:
+            r = client.get("/v1/datasets", headers=headers)
+            if r.status_code == 200:
+                data = r.json()
+                datasets = data.get("data", []) or []
+                names = [d.get("name") for d in datasets]
+                # 能列出知识库即代表鉴权通过、可读取 —— 视为可用。
+                return {
+                    "connected": True,
+                    "usable": True,
+                    "status": "connected",
+                    "detail": f"Dify 连接成功，发现 {len(datasets)} 个知识库: {', '.join([n for n in names if n]) or '无'}",
+                    "datasets": [{"id": d.get("id"), "name": d.get("name")} for d in datasets],
+                }
+            elif r.status_code == 401:
+                return {
+                    "connected": True,
+                    "usable": False,
+                    "status": "auth_error",
+                    "detail": "已连通 Dify，但 API Key 鉴权失败(401)，请检查 API Key。",
+                    "datasets": [],
+                }
+            else:
+                return {
+                    "connected": False,
+                    "usable": False,
+                    "status": "error",
+                    "detail": f"Dify 返回状态码 {r.status_code}",
+                    "datasets": [],
+                }
+    except httpx.TimeoutException:
+        return {"connected": False, "usable": False, "status": "timeout", "detail": "连接超时", "datasets": []}
+    except Exception as e:
+        return {"connected": False, "usable": False, "status": "error", "detail": str(e), "datasets": []}
+
+
+def resolve_dataset_id(api_url: str, api_key: str, dataset_name: str = None) -> str:
+    """解析 dataset id（优先按名称匹配，否则取列表第一个），带缓存。"""
+    base = _normalize_base(api_url)
+    headers = {"Authorization": f"Bearer {api_key}"}
+    cache_key = f"{base}|{api_key}|{dataset_name}"
+    now = time.time()
+    with _dataset_cache_lock:
+        c = _dataset_cache.get(cache_key)
+        if c and now - c[1] < _DATASET_CACHE_TTL:
+            return c[0]
+    ds_id = None
+    try:
+        with httpx.Client(base_url=base, timeout=15, verify=False) as client:
+            r = client.get("/v1/datasets", headers=headers)
+            if r.status_code == 200:
+                datasets = r.json().get("data", []) or []
+                if dataset_name:
+                    match = [d for d in datasets if d.get("name") == dataset_name]
+                    if match:
+                        ds_id = match[0].get("id")
+                if not ds_id and datasets:
+                    ds_id = datasets[0].get("id")
+    except Exception as e:
+        logger.error(f"resolve_dataset_id failed: {e}")
+    if ds_id:
+        with _dataset_cache_lock:
+            _dataset_cache[cache_key] = (ds_id, now)
+    return ds_id
+
+
+def dify_retrieve(api_url, api_key, query, top_k=5, dataset_name=None) -> list:
+    """检索，返回 [{content, score, source, metadata}]。"""
+    ds_id = resolve_dataset_id(api_url, api_key, dataset_name)
+    if not ds_id:
+        return []
+    base = _normalize_base(api_url)
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+    payload = {
+        "query": query,
+        "retrieval_model": {
+            "search_method": "semantic_search",
+            "reranking_enable": False,
+            "score_threshold_enabled": False,
+            "top_k": top_k,
+        },
+    }
+    try:
+        with httpx.Client(base_url=base, timeout=30, verify=False) as client:
+            r = client.post(f"/v1/datasets/{ds_id}/retrieve", headers=headers, json=payload)
+            if r.status_code == 200:
+                data = r.json()
+                out = []
+                for item in data.get("records", []):
+                    seg = item.get("segment", {})
+                    out.append({
+                        "content": seg.get("content", ""),
+                        "score": float(item.get("score", 0.5)),
+                        "source": f"Dify: {seg.get('document_id', '')}",
+                        "metadata": item,
+                    })
+                return out
+            else:
+                logger.warning(f"Dify retrieve returned {r.status_code}: {r.text[:200]}")
+                return []
+    except Exception as e:
+        logger.error(f"Dify retrieve error: {e}")
+        return []
+
+
+def dify_ingest_text(api_url, api_key, text, name=None, dataset_name=None) -> bool:
+    """文本入库（create-by-text）。返回是否成功。"""
+    if not text or not text.strip():
+        return False
+    ds_id = resolve_dataset_id(api_url, api_key, dataset_name)
+    if not ds_id:
+        return False
+    base = _normalize_base(api_url)
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+    payload = {
+        "name": name or f"doc-{int(time.time())}",
+        "text": text,
+        "indexing_technique": "high_quality",
+    }
+    try:
+        with httpx.Client(base_url=base, timeout=60, verify=False) as client:
+            r = client.post(f"/v1/datasets/{ds_id}/document/create-by-text", headers=headers, json=payload)
+            if r.status_code in (200, 201):
+                return True
+            else:
+                logger.warning(f"Dify ingest text returned {r.status_code}: {r.text[:200]}")
+                return False
+    except Exception as e:
+        logger.error(f"Dify ingest text error: {e}")
+        return False
+
+
+async def aingest_text(api_url, api_key, text, name=None, dataset_name=None) -> bool:
+    """异步文本入库（供 chat 实时落库使用）。"""
+    if not text or not text.strip():
+        return False
+    ds_id = resolve_dataset_id(api_url, api_key, dataset_name)
+    if not ds_id:
+        return False
+    base = _normalize_base(api_url)
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+    payload = {
+        "name": name or f"doc-{int(time.time())}",
+        "text": text,
+        "indexing_technique": "high_quality",
+    }
+    try:
+        async with httpx.AsyncClient(base_url=base, timeout=60, verify=False) as client:
+            r = await client.post(f"/v1/datasets/{ds_id}/document/create-by-text", headers=headers, json=payload)
+            if r.status_code in (200, 201):
+                return True
+            else:
+                logger.warning(f"Dify aingest returned {r.status_code}: {r.text[:200]}")
+                return False
+    except Exception as e:
+        logger.error(f"Dify aingest error: {e}")
+        return False
+
+
+def get_dify_kb_context(db, query: str, top_k: int = 5) -> str:
+    """读取已启用的 Dify 外部服务，检索并返回拼接后的上下文文本（供 chat 注入）。"""
+    from sqlalchemy import text as _text
+    try:
+        result = db.execute(_text(
+            "SELECT api_url, api_key, config FROM external_kb_services "
+            "WHERE service_type='dify' AND is_enabled=1 AND status='connected' LIMIT 1"
+        ))
+        row = result.fetchone()
+        if not row:
+            return ""
+        api_url, api_key, config_raw = row
+        dataset_name = None
+        if config_raw:
+            try:
+                cfg = config_raw if isinstance(config_raw, dict) else __import__("json").loads(config_raw)
+                dataset_name = cfg.get("dataset_name")
+            except Exception:
+                pass
+        hits = dify_retrieve(api_url, api_key, query, top_k=top_k, dataset_name=dataset_name)
+        if not hits:
+            return ""
+        parts = []
+        for i, h in enumerate(hits, 1):
+            content = (h.get("content") or "").strip()
+            if content:
+                parts.append(f"[{i}] {content}")
+        return "\n\n".join(parts)
+    except Exception as e:
+        logger.error(f"get_dify_kb_context error: {e}")
+        return ""

backend/python/app/kb_router.py

@@ -1,7 +1,7 @@
 """
 知识库路由层
 ============
-将检索请求智能路由到本地BM25、向量检索、或外部KB服务(FastGPT/Dify/RAGFlow)。
+将检索请求智能路由到本地BM25、向量检索、或外部KB服务(Dify)。
 根据知识库配置自动选择最优检索策略。
 
 使用方式:
@@ -9,17 +9,14 @@
     results = await kb_router_search(query, kb_id, top_k=5)
 """
 
-import os
-import json
 import logging
-import httpx
 from typing import List, Dict, Optional
 
 logger = logging.getLogger("aiflow.kb_router")
 
 
 async def _get_external_services(db_session, enabled_only: bool = True) -> List[Dict]:
-    """获取已配置的外部KB服务"""
+    """获取已配置的外部KB服务（仅 dify）"""
     from sqlalchemy import text
     try:
         query = "SELECT * FROM external_kb_services"
@@ -34,123 +31,34 @@ async def _get_external_services(db_session, enabled_only: bool = True) -> List[
         return []
 
 
-async def _search_fastgpt(api_url: str, api_key: str, query: str, top_k: int = 5) -> List[Dict]:
+async def _search_dify(service: Dict, query: str, top_k: int = 5) -> List[Dict]:
-    """搜索 FastGPT 知识库"""
+    """检索 Dify 知识库（走 app.dify_kb 客户端）"""
     try:
-        async with httpx.AsyncClient(timeout=30) as client:
+        from app.dify_kb import dify_retrieve
-            resp = await client.post(
+        api_url = service.get("api_url", "")
-                f"{api_url.rstrip('/')}/api/v1/dataset/search",
+        api_key = service.get("api_key", "")
-                headers={
+        dataset_name = None
-                    "Authorization": f"Bearer {api_key}",
+        cfg = service.get("config")
-                    "Content-Type": "application/json",
+        if isinstance(cfg, str):
-                },
-                json={
-                    "query": query,
-                    "similarity": 0.5,
-                    "limit": top_k,
-                },
-            )
-            if resp.status_code == 200:
-                data = resp.json()
-                results = []
-                for item in data.get("data", {}).get("list", []):
-                    results.append({
-                        "content": item.get("q", item.get("content", "")),
-                        "score": float(item.get("score", 0.5)),
-                        "source": f"FastGPT: {item.get('collectionId', '')}",
-                        "metadata": item,
-                    })
-                return results
-            else:
-                logger.warning(f"FastGPT search returned {resp.status_code}")
-                return []
-    except Exception as e:
-        logger.error(f"FastGPT search error: {e}")
-        return []
-
-
-async def _search_dify(api_url: str, api_key: str, query: str, top_k: int = 5) -> List[Dict]:
-    """搜索 Dify 知识库"""
             try:
-        async with httpx.AsyncClient(timeout=30) as client:
+                import json
-            resp = await client.post(
+                cfg = json.loads(cfg)
-                f"{api_url.rstrip('/')}/v1/datasets/retrieval",
+            except Exception:
-                headers={
+                cfg = None
-                    "Authorization": f"Bearer {api_key}",
+        if isinstance(cfg, dict):
-                    "Content-Type": "application/json",
+            dataset_name = cfg.get("dataset_name")
-                },
+        hits = dify_retrieve(api_url, api_key, query, top_k=top_k, dataset_name=dataset_name)
-                json={
+        return hits
-                    "query": query,
-                    "retrieval_mode": "semantic",
-                    "top_k": top_k,
-                },
-            )
-            if resp.status_code == 200:
-                data = resp.json()
-                results = []
-                for item in data.get("records", []):
-                    segment = item.get("segment", {})
-                    results.append({
-                        "content": segment.get("content", ""),
-                        "score": float(item.get("score", 0.5)),
-                        "source": f"Dify: {segment.get('id', '')}",
-                        "metadata": item,
-                    })
-                return results
-            else:
-                logger.warning(f"Dify search returned {resp.status_code}")
-                return []
     except Exception as e:
         logger.error(f"Dify search error: {e}")
         return []
 
 
-async def _search_ragflow(api_url: str, api_key: str, query: str, top_k: int = 5) -> List[Dict]:
-    """搜索 RAGFlow 知识库"""
-    try:
-        async with httpx.AsyncClient(timeout=30) as client:
-            resp = await client.post(
-                f"{api_url.rstrip('/')}/api/v1/retrieval",
-                headers={
-                    "Authorization": f"Bearer {api_key}",
-                    "Content-Type": "application/json",
-                },
-                json={
-                    "question": query,
-                    "top_k": top_k,
-                },
-            )
-            if resp.status_code == 200:
-                data = resp.json()
-                results = []
-                for item in data.get("data", {}).get("chunks", []):
-                    results.append({
-                        "content": item.get("content", ""),
-                        "score": float(item.get("similarity", 0.5)),
-                        "source": f"RAGFlow: {item.get('document_id', '')}",
-                        "metadata": item,
-                    })
-                return results
-            else:
-                logger.warning(f"RAGFlow search returned {resp.status_code}")
-                return []
-    except Exception as e:
-        logger.error(f"RAGFlow search error: {e}")
-        return []
-
-
 async def _search_external(service: Dict, query: str, top_k: int = 5) -> List[Dict]:
-    """路由到对应的外部KB服务"""
+    """路由到对应的外部KB服务（当前仅支持 dify）"""
     service_type = service.get("service_type", "")
-    api_url = service.get("api_url", "")
+    if service_type == "dify":
-    api_key = service.get("api_key", "")
+        return await _search_dify(service, query, top_k)
-
-    if service_type == "fastgpt":
-        return await _search_fastgpt(api_url, api_key, query, top_k)
-    elif service_type == "dify":
-        return await _search_dify(api_url, api_key, query, top_k)
-    elif service_type == "ragflow":
-        return await _search_ragflow(api_url, api_key, query, top_k)
     else:
         logger.warning(f"Unknown external KB type: {service_type}")
         return []
@@ -171,7 +79,7 @@ async def kb_router_search(
     检索策略:
     1. 本地 BM25 检索 (MySQL, 已有)
     2. 向量语义检索 (PostgreSQL+pgvector, 新增)
-    3. 外部KB检索 (FastGPT/Dify/RAGFlow, 按配置)
+    3. 外部KB检索 (Dify, 按配置)
 
     融合方式: Reciprocal Rank Fusion (RRF)
 
@@ -218,7 +126,7 @@ async def kb_router_search(
         except Exception as e:
             logger.warning(f"Vector search failed: {e}")
 
-    # 3. 外部KB检索
+    # 3. 外部KB检索 (Dify)
     if use_external and db_session:
         services = await _get_external_services(db_session)
         for service in services:
@@ -228,7 +136,7 @@ async def kb_router_search(
                     all_results.append({
                         "content": item.get("content", ""),
                         "score": item.get("score", 0.5),
-                        "source": item.get("source", f"external_{service['service_type']}"),
+                        "source": item.get("source", "external_dify"),
                         "rank": rank + 1,
                         "metadata": item.get("metadata", {}),
                     })

frontend/src/pages/admin/Knowledge.tsx

@@ -66,7 +66,10 @@ const Knowledge: React.FC = () => {
   // External Services
   const [extServices, setExtServices] = useState<ExternalService[]>([]);
   const [showExtModal, setShowExtModal] = useState(false);
-  const [extForm, setExtForm] = useState({ name: '', service_type: 'fastgpt', api_url: '', api_key: '' });
+  const [extForm, setExtForm] = useState({ name: 'Dify 知识库', service_type: 'dify', api_url: 'https://dify.qihuyun.com/v1', api_key: '', dataset_name: 'qihuai' });
+  const [extMsg, setExtMsg] = useState('');
+  const [ingesting, setIngesting] = useState(false);
+  const [ingestResult, setIngestResult] = useState<any>(null);
 
   // KB Modal
   const [showModal, setShowModal] = useState(false);
@@ -210,7 +213,19 @@ const Knowledge: React.FC = () => {
   };
 
   const handleCreateExtService = async () => {
-    try { const res = await adminAPI.createAdminKnowledgeExtService(extForm); if (res.code === 0) { setShowExtModal(false); setExtForm({ name: '', service_type: 'fastgpt', api_url: '', api_key: '' }); fetchExtServices(); } } catch (e) {}
+    try {
+      const payload = {
+        ...extForm,
+        config: { dataset_name: extForm.dataset_name || 'qihuai' },
+      };
+      const res = await adminAPI.createAdminKnowledgeExtService(payload);
+      if (res.code === 0) {
+        setShowExtModal(false);
+        setExtForm({ name: 'Dify 知识库', service_type: 'dify', api_url: 'https://dify.qihuyun.com/v1', api_key: '', dataset_name: 'qihuai' });
+        setExtMsg('');
+        fetchExtServices();
+      }
+    } catch (e: any) { setExtMsg(`添加失败: ${e?.message || '未知错误'}`); }
   };
 
   const handleDeleteExtService = async (id: string) => {
@@ -219,7 +234,38 @@ const Knowledge: React.FC = () => {
   };
 
   const handleCheckExtStatus = async (id: string) => {
-    try { const res = await adminAPI.checkAdminKnowledgeExtStatus(id); if (res.code === 0) { alert(`连接状态: ${res.data.connected ? '已连接' : '未连接'}\n详情: ${res.data.detail || '无'}`); fetchExtServices(); } } catch (e) {}
+    try {
+      const res = await adminAPI.checkAdminKnowledgeExtStatus(id);
+      if (res.code === 0) {
+        const d = res.data || {};
+        const statusText = d.connected ? (d.usable ? '可通讯且可用 ✅' : '可通讯但不可用 ⚠️') : '不可通讯 ❌';
+        alert(`连接状态: ${statusText}\n\n详情: ${d.detail || '无'}`);
+        fetchExtServices();
+      }
+    } catch (e) {}
+  };
+
+  const handleIngestPlatform = async () => {
+    if (!confirm('将把平台「AI聊天回答 + 用户上传文档」中有价值的内容批量入库到 Dify 知识库（量化后可被 AI 聊天检索复用）。确认执行？')) return;
+    setIngesting(true); setIngestResult(null);
+    try {
+      const token = localStorage.getItem('aiflow_token');
+      const res = await fetch(`/app/api/admin/knowledge/external/dify/ingest`, {
+        method: 'POST',
+        headers: { 'Authorization': `Bearer ${token}` },
+      });
+      const data = await res.json();
+      if (data.code === 0) {
+        setIngestResult(data.data);
+        setExtMsg(`✅ ${data.message}`);
+      } else {
+        setExtMsg(`❌ ${data.message || '入库失败'}`);
+      }
+    } catch (e: any) {
+      setExtMsg(`❌ 入库失败: ${e?.message || '网络错误'}`);
+    } finally {
+      setIngesting(false);
+    }
   };
 
   const statusBadge = (status: string) => {
@@ -487,7 +533,7 @@ const Knowledge: React.FC = () => {
                     {extServices.map(svc => (
                       <tr key={svc.id}>
                         <td>{svc.name}</td>
-                        <td><span style={{ fontSize: 12, padding: '2px 6px', background: '#F3F4FF', color: '#8B5CF6', borderRadius: 3 }}>{svc.service_type === 'fastgpt' ? 'FastGPT' : svc.service_type === 'dify' ? 'Dify' : 'RAGFlow'}</span></td>
+                        <td><span style={{ fontSize: 12, padding: '2px 6px', background: '#F3F4FF', color: '#8B5CF6', borderRadius: 3 }}>Dify</span></td>
                         <td style={{ fontSize: 11, maxWidth: 200, overflow: 'hidden', textOverflow: 'ellipsis' }}>{svc.api_url}</td>
                         <td>{statusBadge(svc.status)}</td>
                         <td>
@@ -501,17 +547,36 @@ const Knowledge: React.FC = () => {
               </div>
             )}
             <div style={{ borderTop: '1px solid #E5E7EB', paddingTop: 16 }}>
-              <h4 style={{ margin: '0 0 12px 0', fontSize: 14 }}>添加新服务</h4>
+              <h4 style={{ margin: '0 0 12px 0', fontSize: 14 }}>配置 Dify 知识库</h4>
               <div style={{ display: 'grid', gridTemplateColumns: '1fr 1fr', gap: 12 }}>
-                <div><label style={labelStyle}>名称</label><input value={extForm.name} onChange={e => setExtForm({ ...extForm, name: e.target.value })} placeholder="如: 生产环境 FastGPT" style={fullInput} /></div>
+                <div><label style={labelStyle}>名称</label><input value={extForm.name} onChange={e => setExtForm({ ...extForm, name: e.target.value })} placeholder="如: 生产环境 Dify" style={fullInput} /></div>
-                <div><label style={labelStyle}>类型</label><select value={extForm.service_type} onChange={e => setExtForm({ ...extForm, service_type: e.target.value })} style={fullInput}><option value="fastgpt">FastGPT</option><option value="dify">Dify</option><option value="ragflow">RAGFlow</option></select></div>
+                <div><label style={labelStyle}>知识库名称(Dataset)</label><input value={extForm.dataset_name} onChange={e => setExtForm({ ...extForm, dataset_name: e.target.value })} placeholder="如: qihuai" style={fullInput} /></div>
               </div>
-              <div style={{ marginTop: 12 }}><label style={labelStyle}>API URL</label><input value={extForm.api_url} onChange={e => setExtForm({ ...extForm, api_url: e.target.value })} placeholder="https://your-instance.com/api/v1" style={fullInput} /></div>
+              <div style={{ marginTop: 12 }}><label style={labelStyle}>API URL</label><input value={extForm.api_url} onChange={e => setExtForm({ ...extForm, api_url: e.target.value })} placeholder="https://dify.qihuyun.com/v1" style={fullInput} /></div>
-              <div style={{ marginTop: 12 }}><label style={labelStyle}>API Key</label><input type="password" value={extForm.api_key} onChange={e => setExtForm({ ...extForm, api_key: e.target.value })} placeholder="Bearer Token 或 API Key" style={fullInput} /></div>
+              <div style={{ marginTop: 12 }}><label style={labelStyle}>API Key（dataset 密钥）</label><input type="password" value={extForm.api_key} onChange={e => setExtForm({ ...extForm, api_key: e.target.value })} placeholder="dataset-xxxx" style={fullInput} /></div>
-              <div style={{ display: 'flex', gap: 12, marginTop: 16 }}>
+
-                <button onClick={handleCreateExtService} style={btnStyle('#8B5CF6')}>添加服务</button>
+              {extMsg && (
+                <div style={{ marginTop: 12, fontSize: 13, color: extMsg.startsWith('✅') ? '#16A34A' : '#DC2626', padding: '6px 12px', background: extMsg.startsWith('✅') ? '#DCFCE7' : '#FEE2E2', borderRadius: 6 }}>
+                  {extMsg}
+                </div>
+              )}
+
+              <div style={{ display: 'flex', gap: 12, marginTop: 16, flexWrap: 'wrap' }}>
+                <button onClick={handleCreateExtService} style={btnStyle('#8B5CF6')}>保存并测试连接</button>
+                <button onClick={handleIngestPlatform} disabled={ingesting} style={btnStyle(ingesting ? '#D1D5DB' : '#219B5E')}>
+                  {ingesting ? '入库中...' : '入库平台内容'}
+                </button>
                 <button onClick={() => setShowExtModal(false)} style={btnStyle('#6B7280')}>关闭</button>
               </div>
+
+              {ingestResult && (
+                <div style={{ marginTop: 12, fontSize: 13, padding: 12, background: '#F0FDF4', border: '1px solid #BBF7D0', borderRadius: 8 }}>
+                  <div style={{ fontWeight: 600, marginBottom: 6 }}>入库结果</div>
+                  <div>AI 聊天回答：<b>{ingestResult.chat_answers}</b> 条</div>
+                  <div>用户上传文档：<b>{ingestResult.user_docs}</b> 条</div>
+                  <div>跳过(过短/空)：{ingestResult.skipped} 条 · 失败：{ingestResult.failed} 条</div>
+                </div>
+              )}
             </div>
           </div>
         </div>