fix: margin in rerank switch

# Conflicts:
#	api/poetry.lock

api/configs/middleware/__init__.py

@@ -5,6 +5,7 @@ from pydantic import Field, NonNegativeInt, PositiveFloat, PositiveInt, computed
 from pydantic_settings import BaseSettings
 
 from configs.middleware.cache.redis_config import RedisConfig
+from configs.middleware.external.bedrock_config import BedrockConfig
 from configs.middleware.storage.aliyun_oss_storage_config import AliyunOSSStorageConfig
 from configs.middleware.storage.amazon_s3_storage_config import S3StorageConfig
 from configs.middleware.storage.azure_blob_storage_config import AzureBlobStorageConfig
@@ -222,5 +223,6 @@ class MiddlewareConfig(
     TiDBVectorConfig,
     WeaviateConfig,
     ElasticsearchConfig,
+    BedrockConfig,
 ):
     pass

api/configs/middleware/external/bedrock_config.py

@@ -0,0 +1,20 @@
+from typing import Optional
+
+from pydantic import Field
+from pydantic_settings import BaseSettings
+
+
+class BedrockConfig(BaseSettings):
+    """
+    bedrock configs
+    """
+
+    AWS_SECRET_ACCESS_KEY: Optional[str] = Field(
+        description="AWS secret access key",
+        default=None,
+    )
+
+    AWS_ACCESS_KEY_ID: Optional[str] = Field(
+        description="AWS secret access id",
+        default=None,
+    )

api/controllers/console/__init__.py

@@ -37,7 +37,17 @@ from .auth import activate, data_source_bearer_auth, data_source_oauth, forgot_p
 from .billing import billing
 
 # Import datasets controllers
-from .datasets import data_source, datasets, datasets_document, datasets_segments, file, hit_testing, website
+from .datasets import (
+    data_source,
+    datasets,
+    datasets_document,
+    datasets_segments,
+    external,
+    file,
+    hit_testing,
+    test_external,
+    website,
+)
 
 # Import explore controllers
 from .explore import (

api/controllers/console/datasets/datasets.py

@@ -49,7 +49,7 @@ class DatasetListApi(Resource):
         page = request.args.get("page", default=1, type=int)
         limit = request.args.get("limit", default=20, type=int)
         ids = request.args.getlist("ids")
-        provider = request.args.get("provider", default="vendor")
+        # provider = request.args.get("provider", default="vendor")
         search = request.args.get("keyword", default=None, type=str)
         tag_ids = request.args.getlist("tag_ids")
 
@@ -57,7 +57,7 @@ class DatasetListApi(Resource):
             datasets, total = DatasetService.get_datasets_by_ids(ids, current_user.current_tenant_id)
         else:
             datasets, total = DatasetService.get_datasets(
-                page, limit, provider, current_user.current_tenant_id, current_user, search, tag_ids
+                page, limit, current_user.current_tenant_id, current_user, search, tag_ids
             )
 
         # check embedding setting
@@ -110,6 +110,26 @@ class DatasetListApi(Resource):
             nullable=True,
             help="Invalid indexing technique.",
         )
+        parser.add_argument(
+            "external_knowledge_api_id",
+            type=str,
+            nullable=True,
+            required=False,
+        )
+        parser.add_argument(
+            "provider",
+            type=str,
+            nullable=True,
+            choices=Dataset.PROVIDER_LIST,
+            required=False,
+            default="vendor",
+        )
+        parser.add_argument(
+            "external_knowledge_id",
+            type=str,
+            nullable=True,
+            required=False,
+        )
         args = parser.parse_args()
 
         # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
@@ -123,6 +143,9 @@ class DatasetListApi(Resource):
                 indexing_technique=args["indexing_technique"],
                 account=current_user,
                 permission=DatasetPermissionEnum.ONLY_ME,
+                provider=args["provider"],
+                external_knowledge_api_id=args["external_knowledge_api_id"],
+                external_knowledge_id=args["external_knowledge_id"],
             )
         except services.errors.dataset.DatasetNameDuplicateError:
             raise DatasetNameDuplicateError()
@@ -211,6 +234,33 @@ class DatasetApi(Resource):
         )
         parser.add_argument("retrieval_model", type=dict, location="json", help="Invalid retrieval model.")
         parser.add_argument("partial_member_list", type=list, location="json", help="Invalid parent user list.")
+
+        parser.add_argument(
+            "external_retrieval_model",
+            type=dict,
+            required=False,
+            nullable=True,
+            location="json",
+            help="Invalid external retrieval model.",
+        )
+
+        parser.add_argument(
+            "external_knowledge_id",
+            type=str,
+            required=False,
+            nullable=True,
+            location="json",
+            help="Invalid external knowledge id.",
+        )
+
+        parser.add_argument(
+            "external_knowledge_api_id",
+            type=str,
+            required=False,
+            nullable=True,
+            location="json",
+            help="Invalid external knowledge api id.",
+        )
         args = parser.parse_args()
         data = request.get_json()
 

api/controllers/console/datasets/external.py

@@ -0,0 +1,282 @@
+from flask import request
+from flask_login import current_user
+from flask_restful import Resource, marshal, reqparse
+from werkzeug.exceptions import Forbidden, InternalServerError, NotFound
+
+import services
+from controllers.console import api
+from controllers.console.app.error import ProviderNotInitializeError
+from controllers.console.datasets.error import DatasetNameDuplicateError
+from controllers.console.setup import setup_required
+from controllers.console.wraps import account_initialization_required
+from fields.dataset_fields import dataset_detail_fields
+from libs.login import login_required
+from services.dataset_service import DatasetService
+from services.external_knowledge_service import ExternalDatasetService
+from services.hit_testing_service import HitTestingService
+
+
+def _validate_name(name):
+    if not name or len(name) < 1 or len(name) > 100:
+        raise ValueError("Name must be between 1 to 100 characters.")
+    return name
+
+
+def _validate_description_length(description):
+    if description and len(description) > 400:
+        raise ValueError("Description cannot exceed 400 characters.")
+    return description
+
+
+class ExternalApiTemplateListApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self):
+        page = request.args.get("page", default=1, type=int)
+        limit = request.args.get("limit", default=20, type=int)
+        search = request.args.get("keyword", default=None, type=str)
+
+        external_knowledge_apis, total = ExternalDatasetService.get_external_knowledge_apis(
+            page, limit, current_user.current_tenant_id, search
+        )
+        response = {
+            "data": [item.to_dict() for item in external_knowledge_apis],
+            "has_more": len(external_knowledge_apis) == limit,
+            "limit": limit,
+            "total": total,
+            "page": page,
+        }
+        return response, 200
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def post(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument(
+            "name",
+            nullable=False,
+            required=True,
+            help="Name is required. Name must be between 1 to 100 characters.",
+            type=_validate_name,
+        )
+        parser.add_argument(
+            "settings",
+            type=dict,
+            location="json",
+            nullable=False,
+            required=True,
+        )
+        args = parser.parse_args()
+
+        ExternalDatasetService.validate_api_list(args["settings"])
+
+        # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
+        if not current_user.is_dataset_editor:
+            raise Forbidden()
+
+        try:
+            external_knowledge_api = ExternalDatasetService.create_external_knowledge_api(
+                tenant_id=current_user.current_tenant_id, user_id=current_user.id, args=args
+            )
+        except services.errors.dataset.DatasetNameDuplicateError:
+            raise DatasetNameDuplicateError()
+
+        return external_knowledge_api.to_dict(), 201
+
+
+class ExternalApiTemplateApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self, external_knowledge_api_id):
+        external_knowledge_api_id = str(external_knowledge_api_id)
+        external_knowledge_api = ExternalDatasetService.get_external_knowledge_api(external_knowledge_api_id)
+        if external_knowledge_api is None:
+            raise NotFound("API template not found.")
+
+        return external_knowledge_api.to_dict(), 200
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def patch(self, external_knowledge_api_id):
+        external_knowledge_api_id = str(external_knowledge_api_id)
+
+        parser = reqparse.RequestParser()
+        parser.add_argument(
+            "name",
+            nullable=False,
+            required=True,
+            help="type is required. Name must be between 1 to 100 characters.",
+            type=_validate_name,
+        )
+        parser.add_argument(
+            "settings",
+            type=dict,
+            location="json",
+            nullable=False,
+            required=True,
+        )
+        args = parser.parse_args()
+        ExternalDatasetService.validate_api_list(args["settings"])
+
+        external_knowledge_api = ExternalDatasetService.update_external_knowledge_api(
+            tenant_id=current_user.current_tenant_id,
+            user_id=current_user.id,
+            external_knowledge_api_id=external_knowledge_api_id,
+            args=args,
+        )
+
+        return external_knowledge_api.to_dict(), 200
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def delete(self, external_knowledge_api_id):
+        external_knowledge_api_id = str(external_knowledge_api_id)
+
+        # The role of the current user in the ta table must be admin, owner, or editor
+        if not current_user.is_editor or current_user.is_dataset_operator:
+            raise Forbidden()
+
+        ExternalDatasetService.delete_external_knowledge_api(current_user.current_tenant_id, external_knowledge_api_id)
+        return {"result": "success"}, 200
+
+
+class ExternalApiUseCheckApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self, external_knowledge_api_id):
+        external_knowledge_api_id = str(external_knowledge_api_id)
+
+        external_knowledge_api_is_using, count = ExternalDatasetService.external_knowledge_api_use_check(
+            external_knowledge_api_id
+        )
+        return {"is_using": external_knowledge_api_is_using, "count": count}, 200
+
+
+class ExternalDatasetInitApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def post(self):
+        # The role of the current user in the ta table must be admin, owner, or editor
+        if not current_user.is_editor:
+            raise Forbidden()
+
+        parser = reqparse.RequestParser()
+        parser.add_argument("external_knowledge_api_id", type=str, required=True, nullable=True, location="json")
+        # parser.add_argument('name', nullable=False, required=True,
+        #                     help='name is required. Name must be between 1 to 100 characters.',
+        #                     type=_validate_name)
+        # parser.add_argument('description', type=str, required=True, nullable=True, location='json')
+        parser.add_argument("data_source", type=dict, required=True, nullable=True, location="json")
+        parser.add_argument("process_parameter", type=dict, required=True, nullable=True, location="json")
+
+        args = parser.parse_args()
+
+        # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
+        if not current_user.is_dataset_editor:
+            raise Forbidden()
+
+        # validate args
+        ExternalDatasetService.document_create_args_validate(
+            current_user.current_tenant_id, args["external_knowledge_api_id"], args["process_parameter"]
+        )
+
+        try:
+            dataset, documents, batch = ExternalDatasetService.init_external_dataset(
+                tenant_id=current_user.current_tenant_id,
+                user_id=current_user.id,
+                args=args,
+            )
+        except Exception as ex:
+            raise ProviderNotInitializeError(ex.description)
+        response = {"dataset": dataset, "documents": documents, "batch": batch}
+
+        return response
+
+
+class ExternalDatasetCreateApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def post(self):
+        # The role of the current user in the ta table must be admin, owner, or editor
+        if not current_user.is_editor:
+            raise Forbidden()
+
+        parser = reqparse.RequestParser()
+        parser.add_argument("external_knowledge_api_id", type=str, required=True, nullable=False, location="json")
+        parser.add_argument("external_knowledge_id", type=str, required=True, nullable=False, location="json")
+        parser.add_argument(
+            "name",
+            nullable=False,
+            required=True,
+            help="name is required. Name must be between 1 to 100 characters.",
+            type=_validate_name,
+        )
+        parser.add_argument("description", type=str, required=False, nullable=True, location="json")
+        parser.add_argument("external_retrieval_model", type=dict, required=False, location="json")
+
+        args = parser.parse_args()
+
+        # The role of the current user in the ta table must be admin, owner, or editor, or dataset_operator
+        if not current_user.is_dataset_editor:
+            raise Forbidden()
+
+        try:
+            dataset = ExternalDatasetService.create_external_dataset(
+                tenant_id=current_user.current_tenant_id,
+                user_id=current_user.id,
+                args=args,
+            )
+        except services.errors.dataset.DatasetNameDuplicateError:
+            raise DatasetNameDuplicateError()
+
+        return marshal(dataset, dataset_detail_fields), 201
+
+
+class ExternalKnowledgeHitTestingApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def post(self, dataset_id):
+        dataset_id_str = str(dataset_id)
+        dataset = DatasetService.get_dataset(dataset_id_str)
+        if dataset is None:
+            raise NotFound("Dataset not found.")
+
+        try:
+            DatasetService.check_dataset_permission(dataset, current_user)
+        except services.errors.account.NoPermissionError as e:
+            raise Forbidden(str(e))
+
+        parser = reqparse.RequestParser()
+        parser.add_argument("query", type=str, location="json")
+        parser.add_argument("external_retrieval_model", type=dict, required=False, location="json")
+        args = parser.parse_args()
+
+        HitTestingService.hit_testing_args_check(args)
+
+        try:
+            response = HitTestingService.external_retrieve(
+                dataset=dataset,
+                query=args["query"],
+                account=current_user,
+                external_retrieval_model=args["external_retrieval_model"],
+            )
+
+            return response
+        except Exception as e:
+            raise InternalServerError(str(e))
+
+
+api.add_resource(ExternalKnowledgeHitTestingApi, "/datasets/<uuid:dataset_id>/external-hit-testing")
+api.add_resource(ExternalDatasetCreateApi, "/datasets/external")
+api.add_resource(ExternalApiTemplateListApi, "/datasets/external-knowledge-api")
+api.add_resource(ExternalApiTemplateApi, "/datasets/external-knowledge-api/<uuid:external_knowledge_api_id>")
+api.add_resource(ExternalApiUseCheckApi, "/datasets/external-knowledge-api/<uuid:external_knowledge_api_id>/use-check")

api/controllers/console/datasets/hit_testing.py

@@ -47,6 +47,7 @@ class HitTestingApi(Resource):
         parser = reqparse.RequestParser()
         parser.add_argument("query", type=str, location="json")
         parser.add_argument("retrieval_model", type=dict, required=False, location="json")
+        parser.add_argument("external_retrieval_model", type=dict, required=False, location="json")
         args = parser.parse_args()
 
         HitTestingService.hit_testing_args_check(args)
@@ -57,6 +58,7 @@ class HitTestingApi(Resource):
                 query=args["query"],
                 account=current_user,
                 retrieval_model=args["retrieval_model"],
+                external_retrieval_model=args["external_retrieval_model"],
                 limit=10,
             )
 

api/controllers/console/datasets/test_external.py

@@ -0,0 +1,33 @@
+from flask_restful import Resource, reqparse
+
+from controllers.console import api
+from controllers.console.setup import setup_required
+from controllers.console.wraps import account_initialization_required
+from libs.login import login_required
+from services.external_knowledge_service import ExternalDatasetService
+
+
+class TestExternalApi(Resource):
+    def post(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument("retrieval_setting", nullable=False, required=True, type=dict, location="json")
+        parser.add_argument(
+            "query",
+            nullable=False,
+            required=True,
+            type=str,
+        )
+        parser.add_argument(
+            "knowledge_id",
+            nullable=False,
+            required=True,
+            type=str,
+        )
+        args = parser.parse_args()
+        result = ExternalDatasetService.test_external_knowledge_retrieval(
+            args["retrieval_setting"], args["query"], args["knowledge_id"]
+        )
+        return result, 200
+
+
+api.add_resource(TestExternalApi, "/retrieval")

api/controllers/service_api/dataset/dataset.py

@@ -28,11 +28,11 @@ class DatasetListApi(DatasetApiResource):
 
         page = request.args.get("page", default=1, type=int)
         limit = request.args.get("limit", default=20, type=int)
-        provider = request.args.get("provider", default="vendor")
+        # provider = request.args.get("provider", default="vendor")
         search = request.args.get("keyword", default=None, type=str)
         tag_ids = request.args.getlist("tag_ids")
 
-        datasets, total = DatasetService.get_datasets(page, limit, provider, tenant_id, current_user, search, tag_ids)
+        datasets, total = DatasetService.get_datasets(page, limit, tenant_id, current_user, search, tag_ids)
         # check embedding setting
         provider_manager = ProviderManager()
         configurations = provider_manager.get_configurations(tenant_id=current_user.current_tenant_id)
@@ -82,6 +82,26 @@ class DatasetListApi(DatasetApiResource):
             required=False,
             nullable=False,
         )
+        parser.add_argument(
+            "external_knowledge_api_id",
+            type=str,
+            nullable=True,
+            required=False,
+            default="_validate_name",
+        )
+        parser.add_argument(
+            "provider",
+            type=str,
+            nullable=True,
+            required=False,
+            default="vendor",
+        )
+        parser.add_argument(
+            "external_knowledge_id",
+            type=str,
+            nullable=True,
+            required=False,
+        )
         args = parser.parse_args()
 
         try:
@@ -91,6 +111,9 @@ class DatasetListApi(DatasetApiResource):
                 indexing_technique=args["indexing_technique"],
                 account=current_user,
                 permission=args["permission"],
+                provider=args["provider"],
+                external_knowledge_api_id=args["external_knowledge_api_id"],
+                external_knowledge_id=args["external_knowledge_id"],
             )
         except services.errors.dataset.DatasetNameDuplicateError:
             raise DatasetNameDuplicateError()

api/core/callback_handler/index_tool_callback_handler.py

@@ -59,7 +59,7 @@ class DatasetIndexToolCallbackHandler:
             for item in resource:
                 dataset_retriever_resource = DatasetRetrieverResource(
                     message_id=self._message_id,
-                    position=item.get("position"),
+                    position=item.get("position") or 0,
                     dataset_id=item.get("dataset_id"),
                     dataset_name=item.get("dataset_name"),
                     document_id=item.get("document_id"),

api/core/rag/datasource/retrieval_service.py

@@ -10,6 +10,7 @@ from core.rag.rerank.constants.rerank_mode import RerankMode
 from core.rag.retrieval.retrieval_methods import RetrievalMethod
 from extensions.ext_database import db
 from models.dataset import Dataset
+from services.external_knowledge_service import ExternalDatasetService
 
 default_retrieval_model = {
     "search_method": RetrievalMethod.SEMANTIC_SEARCH.value,
@@ -34,6 +35,9 @@ class RetrievalService:
         weights: Optional[dict] = None,
     ):
         dataset = db.session.query(Dataset).filter(Dataset.id == dataset_id).first()
+        if not dataset:
+            return []
+
         if not dataset or dataset.available_document_count == 0 or dataset.available_segment_count == 0:
             return []
         all_documents = []
@@ -108,6 +112,16 @@ class RetrievalService:
             )
         return all_documents
 
+    @classmethod
+    def external_retrieve(cls, dataset_id: str, query: str, external_retrieval_model: Optional[dict] = None):
+        dataset = db.session.query(Dataset).filter(Dataset.id == dataset_id).first()
+        if not dataset:
+            return []
+        all_documents = ExternalDatasetService.fetch_external_knowledge_retrieval(
+            dataset.tenant_id, dataset_id, query, external_retrieval_model
+        )
+        return all_documents
+
     @classmethod
     def keyword_search(
         cls, flask_app: Flask, dataset_id: str, query: str, top_k: int, all_documents: list, exceptions: list

api/core/rag/entities/context_entities.py

@@ -0,0 +1,10 @@
+from pydantic import BaseModel
+
+
+class DocumentContext(BaseModel):
+    """
+    Model class for document context.
+    """
+
+    content: str
+    score: float

api/core/rag/models/document.py

@@ -17,6 +17,8 @@ class Document(BaseModel):
     """
     metadata: Optional[dict] = Field(default_factory=dict)
 
+    provider: Optional[str] = "dify"
+
 
 class BaseDocumentTransformer(ABC):
     """Abstract base class for document transformation systems.

api/core/rag/rerank/rerank_model.py

@@ -28,11 +28,16 @@ class RerankModelRunner:
         docs = []
         doc_id = []
         unique_documents = []
-        for document in documents:
+        dify_documents = [item for item in documents if item.provider == "dify"]
+        external_documents = [item for item in documents if item.provider == "external"]
+        for document in dify_documents:
             if document.metadata["doc_id"] not in doc_id:
                 doc_id.append(document.metadata["doc_id"])
                 docs.append(document.page_content)
                 unique_documents.append(document)
+        for document in external_documents:
+            docs.append(document.page_content)
+            unique_documents.append(document)
 
         documents = unique_documents
 
@@ -46,14 +51,10 @@ class RerankModelRunner:
             # format document
             rerank_document = Document(
                 page_content=result.text,
-                metadata={
-                    "doc_id": documents[result.index].metadata["doc_id"],
-                    "doc_hash": documents[result.index].metadata["doc_hash"],
-                    "document_id": documents[result.index].metadata["document_id"],
-                    "dataset_id": documents[result.index].metadata["dataset_id"],
-                    "score": result.score,
-                },
+                metadata=documents[result.index].metadata,
+                provider=documents[result.index].provider,
             )
+            rerank_document.metadata["score"] = result.score
             rerank_documents.append(rerank_document)
 
         return rerank_documents

api/core/rag/retrieval/dataset_retrieval.py

@@ -20,6 +20,7 @@ from core.ops.utils import measure_time
 from core.rag.data_post_processor.data_post_processor import DataPostProcessor
 from core.rag.datasource.keyword.jieba.jieba_keyword_table_handler import JiebaKeywordTableHandler
 from core.rag.datasource.retrieval_service import RetrievalService
+from core.rag.entities.context_entities import DocumentContext
 from core.rag.models.document import Document
 from core.rag.retrieval.retrieval_methods import RetrievalMethod
 from core.rag.retrieval.router.multi_dataset_function_call_router import FunctionCallMultiDatasetRouter
@@ -30,6 +31,7 @@ from core.tools.tool.dataset_retriever.dataset_retriever_tool import DatasetRetr
 from extensions.ext_database import db
 from models.dataset import Dataset, DatasetQuery, DocumentSegment
 from models.dataset import Document as DatasetDocument
+from services.external_knowledge_service import ExternalDatasetService
 
 default_retrieval_model = {
     "search_method": RetrievalMethod.SEMANTIC_SEARCH.value,
@@ -110,7 +112,7 @@ class DatasetRetrieval:
                 continue
 
             # pass if dataset is not available
-            if dataset and dataset.available_document_count == 0:
+            if dataset and dataset.available_document_count == 0 and dataset.provider != "external":
                 continue
 
             available_datasets.append(dataset)
@@ -146,69 +148,93 @@ class DatasetRetrieval:
                 message_id,
             )
 
-        document_score_list = {}
-        for item in all_documents:
-            if item.metadata.get("score"):
-                document_score_list[item.metadata["doc_id"]] = item.metadata["score"]
-
+        dify_documents = [item for item in all_documents if item.provider == "dify"]
+        external_documents = [item for item in all_documents if item.provider == "external"]
         document_context_list = []
-        index_node_ids = [document.metadata["doc_id"] for document in all_documents]
-        segments = DocumentSegment.query.filter(
-            DocumentSegment.dataset_id.in_(dataset_ids),
-            DocumentSegment.completed_at.isnot(None),
-            DocumentSegment.status == "completed",
-            DocumentSegment.enabled == True,
-            DocumentSegment.index_node_id.in_(index_node_ids),
-        ).all()
-
-        if segments:
-            index_node_id_to_position = {id: position for position, id in enumerate(index_node_ids)}
-            sorted_segments = sorted(
-                segments, key=lambda segment: index_node_id_to_position.get(segment.index_node_id, float("inf"))
-            )
-            for segment in sorted_segments:
-                if segment.answer:
-                    document_context_list.append(f"question:{segment.get_sign_content()} answer:{segment.answer}")
-                else:
-                    document_context_list.append(segment.get_sign_content())
-            if show_retrieve_source:
-                context_list = []
-                resource_number = 1
+        retrieval_resource_list = []
+        # deal with external documents
+        for item in external_documents:
+            document_context_list.append(DocumentContext(content=item.page_content, score=item.metadata.get("score")))
+            source = {
+                "dataset_id": item.metadata.get("dataset_id"),
+                "dataset_name": item.metadata.get("dataset_name"),
+                "document_name": item.metadata.get("title"),
+                "data_source_type": "external",
+                "retriever_from": invoke_from.to_source(),
+                "score": item.metadata.get("score"),
+                "content": item.page_content,
+            }
+            retrieval_resource_list.append(source)
+        document_score_list = {}
+        # deal with dify documents
+        if dify_documents:
+            for item in dify_documents:
+                if item.metadata.get("score"):
+                    document_score_list[item.metadata["doc_id"]] = item.metadata["score"]
+
+            index_node_ids = [document.metadata["doc_id"] for document in dify_documents]
+            segments = DocumentSegment.query.filter(
+                DocumentSegment.dataset_id.in_(dataset_ids),
+                DocumentSegment.status == "completed",
+                DocumentSegment.enabled == True,
+                DocumentSegment.index_node_id.in_(index_node_ids),
+            ).all()
+
+            if segments:
+                index_node_id_to_position = {id: position for position, id in enumerate(index_node_ids)}
+                sorted_segments = sorted(
+                    segments, key=lambda segment: index_node_id_to_position.get(segment.index_node_id, float("inf"))
+                )
                 for segment in sorted_segments:
-                    dataset = Dataset.query.filter_by(id=segment.dataset_id).first()
-                    document = DatasetDocument.query.filter(
-                        DatasetDocument.id == segment.document_id,
-                        DatasetDocument.enabled == True,
-                        DatasetDocument.archived == False,
-                    ).first()
-                    if dataset and document:
-                        source = {
-                            "position": resource_number,
-                            "dataset_id": dataset.id,
-                            "dataset_name": dataset.name,
-                            "document_id": document.id,
-                            "document_name": document.name,
-                            "data_source_type": document.data_source_type,
-                            "segment_id": segment.id,
-                            "retriever_from": invoke_from.to_source(),
-                            "score": document_score_list.get(segment.index_node_id, None),
-                        }
-
-                        if invoke_from.to_source() == "dev":
-                            source["hit_count"] = segment.hit_count
-                            source["word_count"] = segment.word_count
-                            source["segment_position"] = segment.position
-                            source["index_node_hash"] = segment.index_node_hash
-                        if segment.answer:
-                            source["content"] = f"question:{segment.content} \nanswer:{segment.answer}"
-                        else:
-                            source["content"] = segment.content
-                        context_list.append(source)
-                    resource_number += 1
-                if hit_callback:
-                    hit_callback.return_retriever_resource_info(context_list)
-
-            return str("\n".join(document_context_list))
+                    if segment.answer:
+                        document_context_list.append(
+                            DocumentContext(
+                                content=f"question:{segment.get_sign_content()} answer:{segment.answer}",
+                                score=document_score_list.get(segment.index_node_id, None),
+                            )
+                        )
+                    else:
+                        document_context_list.append(
+                            DocumentContext(
+                                content=segment.get_sign_content(),
+                                score=document_score_list.get(segment.index_node_id, None),
+                            )
+                        )
+                if show_retrieve_source:
+                    for segment in sorted_segments:
+                        dataset = Dataset.query.filter_by(id=segment.dataset_id).first()
+                        document = DatasetDocument.query.filter(
+                            DatasetDocument.id == segment.document_id,
+                            DatasetDocument.enabled == True,
+                            DatasetDocument.archived == False,
+                        ).first()
+                        if dataset and document:
+                            source = {
+                                "dataset_id": dataset.id,
+                                "dataset_name": dataset.name,
+                                "document_id": document.id,
+                                "document_name": document.name,
+                                "data_source_type": document.data_source_type,
+                                "segment_id": segment.id,
+                                "retriever_from": invoke_from.to_source(),
+                                "score": document_score_list.get(segment.index_node_id, None),
+                            }
+
+                            if invoke_from.to_source() == "dev":
+                                source["hit_count"] = segment.hit_count
+                                source["word_count"] = segment.word_count
+                                source["segment_position"] = segment.position
+                                source["index_node_hash"] = segment.index_node_hash
+                            if segment.answer:
+                                source["content"] = f"question:{segment.content} \nanswer:{segment.answer}"
+                            else:
+                                source["content"] = segment.content
+                            retrieval_resource_list.append(source)
+        if hit_callback and retrieval_resource_list:
+            hit_callback.return_retriever_resource_info(retrieval_resource_list)
+        if document_context_list:
+            document_context_list = sorted(document_context_list, key=lambda x: x.score, reverse=True)
+            return str("\n".join([document_context.content for document_context in document_context_list]))
         return ""
 
     def single_retrieve(
@@ -256,36 +282,58 @@ class DatasetRetrieval:
             # get retrieval model config
             dataset = db.session.query(Dataset).filter(Dataset.id == dataset_id).first()
             if dataset:
-                retrieval_model_config = dataset.retrieval_model or default_retrieval_model
-
-                # get top k
-                top_k = retrieval_model_config["top_k"]
-                # get retrieval method
-                if dataset.indexing_technique == "economy":
-                    retrieval_method = "keyword_search"
-                else:
-                    retrieval_method = retrieval_model_config["search_method"]
-                # get reranking model
-                reranking_model = (
-                    retrieval_model_config["reranking_model"] if retrieval_model_config["reranking_enable"] else None
-                )
-                # get score threshold
-                score_threshold = 0.0
-                score_threshold_enabled = retrieval_model_config.get("score_threshold_enabled")
-                if score_threshold_enabled:
-                    score_threshold = retrieval_model_config.get("score_threshold")
-
-                with measure_time() as timer:
-                    results = RetrievalService.retrieve(
-                        retrieval_method=retrieval_method,
-                        dataset_id=dataset.id,
+                results = []
+                if dataset.provider == "external":
+                    external_documents = ExternalDatasetService.fetch_external_knowledge_retrieval(
+                        tenant_id=dataset.tenant_id,
+                        dataset_id=dataset_id,
                         query=query,
-                        top_k=top_k,
-                        score_threshold=score_threshold,
-                        reranking_model=reranking_model,
-                        reranking_mode=retrieval_model_config.get("reranking_mode", "reranking_model"),
-                        weights=retrieval_model_config.get("weights", None),
+                        external_retrieval_parameters=dataset.retrieval_model,
+                    )
+                    for external_document in external_documents:
+                        document = Document(
+                            page_content=external_document.get("content"),
+                            metadata=external_document.get("metadata"),
+                            provider="external",
+                        )
+                        document.metadata["score"] = external_document.get("score")
+                        document.metadata["title"] = external_document.get("title")
+                        document.metadata["dataset_id"] = dataset_id
+                        document.metadata["dataset_name"] = dataset.name
+                        results.append(document)
+                else:
+                    retrieval_model_config = dataset.retrieval_model or default_retrieval_model
+
+                    # get top k
+                    top_k = retrieval_model_config["top_k"]
+                    # get retrieval method
+                    if dataset.indexing_technique == "economy":
+                        retrieval_method = "keyword_search"
+                    else:
+                        retrieval_method = retrieval_model_config["search_method"]
+                    # get reranking model
+                    reranking_model = (
+                        retrieval_model_config["reranking_model"]
+                        if retrieval_model_config["reranking_enable"]
+                        else None
                     )
+                    # get score threshold
+                    score_threshold = 0.0
+                    score_threshold_enabled = retrieval_model_config.get("score_threshold_enabled")
+                    if score_threshold_enabled:
+                        score_threshold = retrieval_model_config.get("score_threshold")
+
+                    with measure_time() as timer:
+                        results = RetrievalService.retrieve(
+                            retrieval_method=retrieval_method,
+                            dataset_id=dataset.id,
+                            query=query,
+                            top_k=top_k,
+                            score_threshold=score_threshold,
+                            reranking_model=reranking_model,
+                            reranking_mode=retrieval_model_config.get("reranking_mode", "reranking_model"),
+                            weights=retrieval_model_config.get("weights", None),
+                        )
                 self._on_query(query, [dataset_id], app_id, user_from, user_id)
 
                 if results:
@@ -356,7 +404,8 @@ class DatasetRetrieval:
         self, documents: list[Document], message_id: Optional[str] = None, timer: Optional[dict] = None
     ) -> None:
         """Handle retrieval end."""
-        for document in documents:
+        dify_documents = [document for document in documents if document.provider == "dify"]
+        for document in dify_documents:
             query = db.session.query(DocumentSegment).filter(
                 DocumentSegment.index_node_id == document.metadata["doc_id"]
             )
@@ -409,35 +458,54 @@ class DatasetRetrieval:
             if not dataset:
                 return []
 
-            # get retrieval model , if the model is not setting , using default
-            retrieval_model = dataset.retrieval_model or default_retrieval_model
-
-            if dataset.indexing_technique == "economy":
-                # use keyword table query
-                documents = RetrievalService.retrieve(
-                    retrieval_method="keyword_search", dataset_id=dataset.id, query=query, top_k=top_k
+            if dataset.provider == "external":
+                external_documents = ExternalDatasetService.fetch_external_knowledge_retrieval(
+                    tenant_id=dataset.tenant_id,
+                    dataset_id=dataset_id,
+                    query=query,
+                    external_retrieval_parameters=dataset.retrieval_model,
                 )
-                if documents:
-                    all_documents.extend(documents)
+                for external_document in external_documents:
+                    document = Document(
+                        page_content=external_document.get("content"),
+                        metadata=external_document.get("metadata"),
+                        provider="external",
+                    )
+                    document.metadata["score"] = external_document.get("score")
+                    document.metadata["title"] = external_document.get("title")
+                    document.metadata["dataset_id"] = dataset_id
+                    document.metadata["dataset_name"] = dataset.name
+                    all_documents.append(document)
             else:
-                if top_k > 0:
-                    # retrieval source
+                # get retrieval model , if the model is not setting , using default
+                retrieval_model = dataset.retrieval_model or default_retrieval_model
+
+                if dataset.indexing_technique == "economy":
+                    # use keyword table query
                     documents = RetrievalService.retrieve(
-                        retrieval_method=retrieval_model["search_method"],
-                        dataset_id=dataset.id,
-                        query=query,
-                        top_k=retrieval_model.get("top_k") or 2,
-                        score_threshold=retrieval_model.get("score_threshold", 0.0)
-                        if retrieval_model["score_threshold_enabled"]
-                        else 0.0,
-                        reranking_model=retrieval_model.get("reranking_model", None)
-                        if retrieval_model["reranking_enable"]
-                        else None,
-                        reranking_mode=retrieval_model.get("reranking_mode") or "reranking_model",
-                        weights=retrieval_model.get("weights", None),
+                        retrieval_method="keyword_search", dataset_id=dataset.id, query=query, top_k=top_k
                     )
-
-                    all_documents.extend(documents)
+                    if documents:
+                        all_documents.extend(documents)
+                else:
+                    if top_k > 0:
+                        # retrieval source
+                        documents = RetrievalService.retrieve(
+                            retrieval_method=retrieval_model["search_method"],
+                            dataset_id=dataset.id,
+                            query=query,
+                            top_k=retrieval_model.get("top_k") or 2,
+                            score_threshold=retrieval_model.get("score_threshold", 0.0)
+                            if retrieval_model["score_threshold_enabled"]
+                            else 0.0,
+                            reranking_model=retrieval_model.get("reranking_model", None)
+                            if retrieval_model["reranking_enable"]
+                            else None,
+                            reranking_mode=retrieval_model.get("reranking_mode") or "reranking_model",
+                            weights=retrieval_model.get("weights", None),
+                        )
+
+                        all_documents.extend(documents)
 
     def to_dataset_retriever_tool(
         self,

api/core/workflow/nodes/knowledge_retrieval/knowledge_retrieval_node.py

@@ -156,16 +156,34 @@ class KnowledgeRetrievalNode(BaseNode):
                 weights,
                 node_data.multiple_retrieval_config.reranking_enable,
             )
-
-        context_list = []
-        if all_documents:
+        dify_documents = [item for item in all_documents if item.provider == "dify"]
+        external_documents = [item for item in all_documents if item.provider == "external"]
+        retrieval_resource_list = []
+        # deal with external documents
+        for item in external_documents:
+            source = {
+                "metadata": {
+                    "_source": "knowledge",
+                    "dataset_id": item.metadata.get("dataset_id"),
+                    "dataset_name": item.metadata.get("dataset_name"),
+                    "document_name": item.metadata.get("title"),
+                    "data_source_type": "external",
+                    "retriever_from": "workflow",
+                    "score": item.metadata.get("score"),
+                },
+                "title": item.metadata.get("title"),
+                "content": item.page_content,
+            }
+            retrieval_resource_list.append(source)
+        document_score_list = {}
+        # deal with dify documents
+        if dify_documents:
             document_score_list = {}
-            page_number_list = {}
-            for item in all_documents:
+            for item in dify_documents:
                 if item.metadata.get("score"):
                     document_score_list[item.metadata["doc_id"]] = item.metadata["score"]
 
-            index_node_ids = [document.metadata["doc_id"] for document in all_documents]
+            index_node_ids = [document.metadata["doc_id"] for document in dify_documents]
             segments = DocumentSegment.query.filter(
                 DocumentSegment.dataset_id.in_(dataset_ids),
                 DocumentSegment.completed_at.isnot(None),
@@ -186,13 +204,10 @@ class KnowledgeRetrievalNode(BaseNode):
                         Document.enabled == True,
                         Document.archived == False,
                     ).first()
-
-                    resource_number = 1
                     if dataset and document:
                         source = {
                             "metadata": {
                                 "_source": "knowledge",
-                                "position": resource_number,
                                 "dataset_id": dataset.id,
                                 "dataset_name": dataset.name,
                                 "document_id": document.id,
@@ -212,9 +227,14 @@ class KnowledgeRetrievalNode(BaseNode):
                             source["content"] = f"question:{segment.get_sign_content()} \nanswer:{segment.answer}"
                         else:
                             source["content"] = segment.get_sign_content()
-                        context_list.append(source)
-                        resource_number += 1
-        return context_list
+                        retrieval_resource_list.append(source)
+        if retrieval_resource_list:
+            retrieval_resource_list = sorted(retrieval_resource_list, key=lambda x: x.get("score"), reverse=True)
+            position = 1
+            for item in retrieval_resource_list:
+                item["metadata"]["position"] = position
+                position += 1
+        return retrieval_resource_list
 
     @classmethod
     def _extract_variable_selector_to_variable_mapping(

api/fields/dataset_fields.py

@@ -38,9 +38,20 @@ dataset_retrieval_model_fields = {
     "score_threshold_enabled": fields.Boolean,
     "score_threshold": fields.Float,
 }
+external_retrieval_model_fields = {
+    "top_k": fields.Integer,
+    "score_threshold": fields.Float,
+}
 
 tag_fields = {"id": fields.String, "name": fields.String, "type": fields.String}
 
+external_knowledge_info_fields = {
+    "external_knowledge_id": fields.String,
+    "external_knowledge_api_id": fields.String,
+    "external_knowledge_api_name": fields.String,
+    "external_knowledge_api_endpoint": fields.String,
+}
+
 dataset_detail_fields = {
     "id": fields.String,
     "name": fields.String,
@@ -61,6 +72,8 @@ dataset_detail_fields = {
     "embedding_available": fields.Boolean,
     "retrieval_model_dict": fields.Nested(dataset_retrieval_model_fields),
     "tags": fields.List(fields.Nested(tag_fields)),
+    "external_knowledge_info": fields.Nested(external_knowledge_info_fields),
+    "external_retrieval_model": fields.Nested(external_retrieval_model_fields, allow_null=True),
 }
 
 dataset_query_detail_fields = {

api/fields/external_dataset_fields.py

@@ -0,0 +1,11 @@
+from flask_restful import fields
+
+from libs.helper import TimestampField
+
+external_knowledge_api_query_detail_fields = {
+    "id": fields.String,
+    "name": fields.String,
+    "setting": fields.String,
+    "created_by": fields.String,
+    "created_at": TimestampField,
+}

api/migrations/versions/2024_09_24_0922-6af6a521a53e_update_retrieval_resource.py

@@ -0,0 +1,48 @@
+"""update-retrieval-resource
+
+Revision ID: 6af6a521a53e
+Revises: ec3df697ebbb
+Create Date: 2024-09-24 09:22:43.570120
+
+"""
+from alembic import op
+import models as models
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = '6af6a521a53e'
+down_revision = 'd57ba9ebb251'
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    with op.batch_alter_table('dataset_retriever_resources', schema=None) as batch_op:
+        batch_op.alter_column('document_id',
+               existing_type=sa.UUID(),
+               nullable=True)
+        batch_op.alter_column('data_source_type',
+               existing_type=sa.TEXT(),
+               nullable=True)
+        batch_op.alter_column('segment_id',
+               existing_type=sa.UUID(),
+               nullable=True)
+    # ### end Alembic commands ###
+
+
+def downgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    with op.batch_alter_table('dataset_retriever_resources', schema=None) as batch_op:
+        batch_op.alter_column('segment_id',
+               existing_type=sa.UUID(),
+               nullable=False)
+        batch_op.alter_column('data_source_type',
+               existing_type=sa.TEXT(),
+               nullable=False)
+        batch_op.alter_column('document_id',
+               existing_type=sa.UUID(),
+               nullable=False)
+
+    # ### end Alembic commands ###

api/migrations/versions/2024_09_25_0434-33f5fac87f29_external_knowledge_api.py

@@ -0,0 +1,73 @@
+"""external_knowledge_api
+
+Revision ID: 33f5fac87f29
+Revises: 6af6a521a53e
+Create Date: 2024-09-25 04:34:57.249436
+
+"""
+from alembic import op
+import models as models
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = '33f5fac87f29'
+down_revision = '6af6a521a53e'
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table('external_knowledge_apis',
+    sa.Column('id', models.types.StringUUID(), server_default=sa.text('uuid_generate_v4()'), nullable=False),
+    sa.Column('name', sa.String(length=255), nullable=False),
+    sa.Column('description', sa.String(length=255), nullable=False),
+    sa.Column('tenant_id', models.types.StringUUID(), nullable=False),
+    sa.Column('settings', sa.Text(), nullable=True),
+    sa.Column('created_by', models.types.StringUUID(), nullable=False),
+    sa.Column('created_at', sa.DateTime(), server_default=sa.text('CURRENT_TIMESTAMP(0)'), nullable=False),
+    sa.Column('updated_by', models.types.StringUUID(), nullable=True),
+    sa.Column('updated_at', sa.DateTime(), server_default=sa.text('CURRENT_TIMESTAMP(0)'), nullable=False),
+    sa.PrimaryKeyConstraint('id', name='external_knowledge_apis_pkey')
+    )
+    with op.batch_alter_table('external_knowledge_apis', schema=None) as batch_op:
+        batch_op.create_index('external_knowledge_apis_name_idx', ['name'], unique=False)
+        batch_op.create_index('external_knowledge_apis_tenant_idx', ['tenant_id'], unique=False)
+
+    op.create_table('external_knowledge_bindings',
+    sa.Column('id', models.types.StringUUID(), server_default=sa.text('uuid_generate_v4()'), nullable=False),
+    sa.Column('tenant_id', models.types.StringUUID(), nullable=False),
+    sa.Column('external_knowledge_api_id', models.types.StringUUID(), nullable=False),
+    sa.Column('dataset_id', models.types.StringUUID(), nullable=False),
+    sa.Column('external_knowledge_id', sa.Text(), nullable=False),
+    sa.Column('created_by', models.types.StringUUID(), nullable=False),
+    sa.Column('created_at', sa.DateTime(), server_default=sa.text('CURRENT_TIMESTAMP(0)'), nullable=False),
+    sa.Column('updated_by', models.types.StringUUID(), nullable=True),
+    sa.Column('updated_at', sa.DateTime(), server_default=sa.text('CURRENT_TIMESTAMP(0)'), nullable=False),
+    sa.PrimaryKeyConstraint('id', name='external_knowledge_bindings_pkey')
+    )
+    with op.batch_alter_table('external_knowledge_bindings', schema=None) as batch_op:
+        batch_op.create_index('external_knowledge_bindings_dataset_idx', ['dataset_id'], unique=False)
+        batch_op.create_index('external_knowledge_bindings_external_knowledge_api_idx', ['external_knowledge_api_id'], unique=False)
+        batch_op.create_index('external_knowledge_bindings_external_knowledge_idx', ['external_knowledge_id'], unique=False)
+        batch_op.create_index('external_knowledge_bindings_tenant_idx', ['tenant_id'], unique=False)
+
+    # ### end Alembic commands ###
+
+
+def downgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    with op.batch_alter_table('external_knowledge_bindings', schema=None) as batch_op:
+        batch_op.drop_index('external_knowledge_bindings_tenant_idx')
+        batch_op.drop_index('external_knowledge_bindings_external_knowledge_idx')
+        batch_op.drop_index('external_knowledge_bindings_external_knowledge_api_idx')
+        batch_op.drop_index('external_knowledge_bindings_dataset_idx')
+
+    op.drop_table('external_knowledge_bindings')
+    with op.batch_alter_table('external_knowledge_apis', schema=None) as batch_op:
+        batch_op.drop_index('external_knowledge_apis_tenant_idx')
+        batch_op.drop_index('external_knowledge_apis_name_idx')
+
+    op.drop_table('external_knowledge_apis')
+    # ### end Alembic commands ###

api/migrations/versions/fca025d3b60f_add_dataset_retrival_model.py

@@ -1,4 +1,4 @@
-"""add-dataset-retrival-model
+"""add-dataset-retrieval-model
 
 Revision ID: fca025d3b60f
 Revises: b3a09c049e8e

api/models/dataset.py

@@ -38,6 +38,7 @@ class Dataset(db.Model):
     )
 
     INDEXING_TECHNIQUE_LIST = ["high_quality", "economy", None]
+    PROVIDER_LIST = ["vendor", "external", None]
 
     id = db.Column(StringUUID, server_default=db.text("uuid_generate_v4()"))
     tenant_id = db.Column(StringUUID, nullable=False)
@@ -71,6 +72,14 @@ class Dataset(db.Model):
     def index_struct_dict(self):
         return json.loads(self.index_struct) if self.index_struct else None
 
+    @property
+    def external_retrieval_model(self):
+        default_retrieval_model = {
+            "top_k": 2,
+            "score_threshold": 0.0,
+        }
+        return self.retrieval_model or default_retrieval_model
+
     @property
     def created_by_account(self):
         return db.session.get(Account, self.created_by)
@@ -162,6 +171,29 @@ class Dataset(db.Model):
 
         return tags or []
 
+    @property
+    def external_knowledge_info(self):
+        if self.provider != "external":
+            return None
+        external_knowledge_binding = (
+            db.session.query(ExternalKnowledgeBindings).filter(ExternalKnowledgeBindings.dataset_id == self.id).first()
+        )
+        if not external_knowledge_binding:
+            return None
+        external_knowledge_api = (
+            db.session.query(ExternalKnowledgeApis)
+            .filter(ExternalKnowledgeApis.id == external_knowledge_binding.external_knowledge_api_id)
+            .first()
+        )
+        if not external_knowledge_api:
+            return None
+        return {
+            "external_knowledge_id": external_knowledge_binding.external_knowledge_id,
+            "external_knowledge_api_id": external_knowledge_api.id,
+            "external_knowledge_api_name": external_knowledge_api.name,
+            "external_knowledge_api_endpoint": json.loads(external_knowledge_api.settings).get("endpoint", ""),
+        }
+
     @staticmethod
     def gen_collection_name_by_id(dataset_id: str) -> str:
         normalized_dataset_id = dataset_id.replace("-", "_")
@@ -687,3 +719,77 @@ class DatasetPermission(db.Model):
     tenant_id = db.Column(StringUUID, nullable=False)
     has_permission = db.Column(db.Boolean, nullable=False, server_default=db.text("true"))
     created_at = db.Column(db.DateTime, nullable=False, server_default=db.text("CURRENT_TIMESTAMP(0)"))
+
+
+class ExternalKnowledgeApis(db.Model):
+    __tablename__ = "external_knowledge_apis"
+    __table_args__ = (
+        db.PrimaryKeyConstraint("id", name="external_knowledge_apis_pkey"),
+        db.Index("external_knowledge_apis_tenant_idx", "tenant_id"),
+        db.Index("external_knowledge_apis_name_idx", "name"),
+    )
+
+    id = db.Column(StringUUID, nullable=False, server_default=db.text("uuid_generate_v4()"))
+    name = db.Column(db.String(255), nullable=False)
+    description = db.Column(db.String(255), nullable=False)
+    tenant_id = db.Column(StringUUID, nullable=False)
+    settings = db.Column(db.Text, nullable=True)
+    created_by = db.Column(StringUUID, nullable=False)
+    created_at = db.Column(db.DateTime, nullable=False, server_default=db.text("CURRENT_TIMESTAMP(0)"))
+    updated_by = db.Column(StringUUID, nullable=True)
+    updated_at = db.Column(db.DateTime, nullable=False, server_default=db.text("CURRENT_TIMESTAMP(0)"))
+
+    def to_dict(self):
+        return {
+            "id": self.id,
+            "tenant_id": self.tenant_id,
+            "name": self.name,
+            "description": self.description,
+            "settings": self.settings_dict,
+            "dataset_bindings": self.dataset_bindings,
+            "created_by": self.created_by,
+            "created_at": self.created_at.isoformat(),
+        }
+
+    @property
+    def settings_dict(self):
+        try:
+            return json.loads(self.settings) if self.settings else None
+        except JSONDecodeError:
+            return None
+
+    @property
+    def dataset_bindings(self):
+        external_knowledge_bindings = (
+            db.session.query(ExternalKnowledgeBindings)
+            .filter(ExternalKnowledgeBindings.external_knowledge_api_id == self.id)
+            .all()
+        )
+        dataset_ids = [binding.dataset_id for binding in external_knowledge_bindings]
+        datasets = db.session.query(Dataset).filter(Dataset.id.in_(dataset_ids)).all()
+        dataset_bindings = []
+        for dataset in datasets:
+            dataset_bindings.append({"id": dataset.id, "name": dataset.name})
+
+        return dataset_bindings
+
+
+class ExternalKnowledgeBindings(db.Model):
+    __tablename__ = "external_knowledge_bindings"
+    __table_args__ = (
+        db.PrimaryKeyConstraint("id", name="external_knowledge_bindings_pkey"),
+        db.Index("external_knowledge_bindings_tenant_idx", "tenant_id"),
+        db.Index("external_knowledge_bindings_dataset_idx", "dataset_id"),
+        db.Index("external_knowledge_bindings_external_knowledge_idx", "external_knowledge_id"),
+        db.Index("external_knowledge_bindings_external_knowledge_api_idx", "external_knowledge_api_id"),
+    )
+
+    id = db.Column(StringUUID, nullable=False, server_default=db.text("uuid_generate_v4()"))
+    tenant_id = db.Column(StringUUID, nullable=False)
+    external_knowledge_api_id = db.Column(StringUUID, nullable=False)
+    dataset_id = db.Column(StringUUID, nullable=False)
+    external_knowledge_id = db.Column(db.Text, nullable=False)
+    created_by = db.Column(StringUUID, nullable=False)
+    created_at = db.Column(db.DateTime, nullable=False, server_default=db.text("CURRENT_TIMESTAMP(0)"))
+    updated_by = db.Column(StringUUID, nullable=True)
+    updated_at = db.Column(db.DateTime, nullable=False, server_default=db.text("CURRENT_TIMESTAMP(0)"))

api/models/model.py

@@ -1423,10 +1423,10 @@ class DatasetRetrieverResource(db.Model):
     position = db.Column(db.Integer, nullable=False)
     dataset_id = db.Column(StringUUID, nullable=False)
     dataset_name = db.Column(db.Text, nullable=False)
-    document_id = db.Column(StringUUID, nullable=False)
+    document_id = db.Column(StringUUID, nullable=True)
     document_name = db.Column(db.Text, nullable=False)
-    data_source_type = db.Column(db.Text, nullable=False)
-    segment_id = db.Column(StringUUID, nullable=False)
+    data_source_type = db.Column(db.Text, nullable=True)
+    segment_id = db.Column(StringUUID, nullable=True)
     score = db.Column(db.Float, nullable=True)
     content = db.Column(db.Text, nullable=False)
     hit_count = db.Column(db.Integer, nullable=True)

api/poetry.lock

@@ -6644,19 +6644,6 @@ files = [
     {file = "pyarrow-17.0.0-cp312-cp312-win_amd64.whl", hash = "sha256:392bc9feabc647338e6c89267635e111d71edad5fcffba204425a7c8d13610d7"},
     {file = "pyarrow-17.0.0-cp38-cp38-macosx_10_15_x86_64.whl", hash = "sha256:af5ff82a04b2171415f1410cff7ebb79861afc5dae50be73ce06d6e870615204"},
     {file = "pyarrow-17.0.0-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:edca18eaca89cd6382dfbcff3dd2d87633433043650c07375d095cd3517561d8"},
-    {file = "pyarrow-17.0.0-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:7c7916bff914ac5d4a8fe25b7a25e432ff921e72f6f2b7547d1e325c1ad9d155"},
-    {file = "pyarrow-17.0.0-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:f553ca691b9e94b202ff741bdd40f6ccb70cdd5fbf65c187af132f1317de6145"},
-    {file = "pyarrow-17.0.0-cp38-cp38-manylinux_2_28_aarch64.whl", hash = "sha256:0cdb0e627c86c373205a2f94a510ac4376fdc523f8bb36beab2e7f204416163c"},
-    {file = "pyarrow-17.0.0-cp38-cp38-manylinux_2_28_x86_64.whl", hash = "sha256:d7d192305d9d8bc9082d10f361fc70a73590a4c65cf31c3e6926cd72b76bc35c"},
-    {file = "pyarrow-17.0.0-cp38-cp38-win_amd64.whl", hash = "sha256:02dae06ce212d8b3244dd3e7d12d9c4d3046945a5933d28026598e9dbbda1fca"},
-    {file = "pyarrow-17.0.0-cp39-cp39-macosx_10_15_x86_64.whl", hash = "sha256:13d7a460b412f31e4c0efa1148e1d29bdf18ad1411eb6757d38f8fbdcc8645fb"},
-    {file = "pyarrow-17.0.0-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:9b564a51fbccfab5a04a80453e5ac6c9954a9c5ef2890d1bcf63741909c3f8df"},
-    {file = "pyarrow-17.0.0-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:32503827abbc5aadedfa235f5ece8c4f8f8b0a3cf01066bc8d29de7539532687"},
-    {file = "pyarrow-17.0.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:a155acc7f154b9ffcc85497509bcd0d43efb80d6f733b0dc3bb14e281f131c8b"},
-    {file = "pyarrow-17.0.0-cp39-cp39-manylinux_2_28_aarch64.whl", hash = "sha256:dec8d129254d0188a49f8a1fc99e0560dc1b85f60af729f47de4046015f9b0a5"},
-    {file = "pyarrow-17.0.0-cp39-cp39-manylinux_2_28_x86_64.whl", hash = "sha256:a48ddf5c3c6a6c505904545c25a4ae13646ae1f8ba703c4df4a1bfe4f4006bda"},
-    {file = "pyarrow-17.0.0-cp39-cp39-win_amd64.whl", hash = "sha256:42bf93249a083aca230ba7e2786c5f673507fa97bbd9725a1e2754715151a204"},
-    {file = "pyarrow-17.0.0.tar.gz", hash = "sha256:4beca9521ed2c0921c1023e68d097d0299b62c362639ea315572a58f3f50fd28"},
 ]
 
 [package.dependencies]
@@ -10501,4 +10488,4 @@ cffi = ["cffi (>=1.11)"]
 [metadata]
 lock-version = "2.0"
 python-versions = ">=3.10,<3.13"
-content-hash = "69b42bb1ff033f14e199fee8335356275099421d72bbd7037b7a991ea65cae08"
+content-hash = "1f9d36b61528276a0761d87ef4f9fa787b5c1b49ae85b238c86626fa1110e2e8"

api/pyproject.toml

@@ -220,6 +220,7 @@ volcengine-python-sdk = {extras = ["ark"], version = "^1.0.98"}
 oci = "^2.133.0"
 tos = "^2.7.1"
 nomic = "^3.1.2"
+validators = "0.21.0"
 [tool.poetry.group.indriect.dependencies]
 kaleido = "0.2.1"
 rank-bm25 = "~0.2.2"

api/schedule/clean_unused_messages_task.py

@@ -0,0 +1,92 @@
+import datetime
+import time
+
+import click
+from sqlalchemy import func
+from werkzeug.exceptions import NotFound
+
+import app
+from configs import dify_config
+from core.rag.index_processor.index_processor_factory import IndexProcessorFactory
+from extensions.ext_database import db
+from models.dataset import Dataset, DatasetQuery, Document
+
+
+@app.celery.task(queue="dataset")
+def clean_unused_message_task():
+    click.echo(click.style("Start clean unused messages .", fg="green"))
+    clean_days = int(dify_config.CLEAN_DAY_SETTING)
+    start_at = time.perf_counter()
+    thirty_days_ago = datetime.datetime.now() - datetime.timedelta(days=clean_days)
+    page = 1
+    while True:
+        try:
+            # Subquery for counting new documents
+            document_subquery_new = (
+                db.session.query(Document.dataset_id, func.count(Document.id).label("document_count"))
+                .filter(
+                    Document.indexing_status == "completed",
+                    Document.enabled == True,
+                    Document.archived == False,
+                    Document.updated_at > thirty_days_ago,
+                )
+                .group_by(Document.dataset_id)
+                .subquery()
+            )
+
+            # Subquery for counting old documents
+            document_subquery_old = (
+                db.session.query(Document.dataset_id, func.count(Document.id).label("document_count"))
+                .filter(
+                    Document.indexing_status == "completed",
+                    Document.enabled == True,
+                    Document.archived == False,
+                    Document.updated_at < thirty_days_ago,
+                )
+                .group_by(Document.dataset_id)
+                .subquery()
+            )
+
+            # Main query with join and filter
+            datasets = (
+                db.session.query(Dataset)
+                .outerjoin(document_subquery_new, Dataset.id == document_subquery_new.c.dataset_id)
+                .outerjoin(document_subquery_old, Dataset.id == document_subquery_old.c.dataset_id)
+                .filter(
+                    Dataset.created_at < thirty_days_ago,
+                    func.coalesce(document_subquery_new.c.document_count, 0) == 0,
+                    func.coalesce(document_subquery_old.c.document_count, 0) > 0,
+                )
+                .order_by(Dataset.created_at.desc())
+                .paginate(page=page, per_page=50)
+            )
+
+        except NotFound:
+            break
+        if datasets.items is None or len(datasets.items) == 0:
+            break
+        page += 1
+        for dataset in datasets:
+            dataset_query = (
+                db.session.query(DatasetQuery)
+                .filter(DatasetQuery.created_at > thirty_days_ago, DatasetQuery.dataset_id == dataset.id)
+                .all()
+            )
+            if not dataset_query or len(dataset_query) == 0:
+                try:
+                    # remove index
+                    index_processor = IndexProcessorFactory(dataset.doc_form).init_index_processor()
+                    index_processor.clean(dataset, None)
+
+                    # update document
+                    update_params = {Document.enabled: False}
+
+                    Document.query.filter_by(dataset_id=dataset.id).update(update_params)
+                    db.session.commit()
+                    click.echo(click.style("Cleaned unused dataset {} from db success!".format(dataset.id), fg="green"))
+                except Exception as e:
+                    click.echo(
+                        click.style("clean dataset index error: {} {}".format(e.__class__.__name__, str(e)), fg="red")
+                    )
+    end_at = time.perf_counter()
+    click.echo(click.style("Cleaned unused dataset from db success latency: {}".format(end_at - start_at), fg="green"))

api/services/dataset_service.py

@@ -32,6 +32,7 @@ from models.dataset import (
     DatasetQuery,
     Document,
     DocumentSegment,
+    ExternalKnowledgeBindings,
 )
 from models.model import UploadFile
 from models.source import DataSourceOauthBinding
@@ -39,6 +40,7 @@ from services.errors.account import NoPermissionError
 from services.errors.dataset import DatasetNameDuplicateError
 from services.errors.document import DocumentIndexingError
 from services.errors.file import FileNotExistsError
+from services.external_knowledge_service import ExternalDatasetService
 from services.feature_service import FeatureModel, FeatureService
 from services.tag_service import TagService
 from services.vector_service import VectorService
@@ -56,10 +58,8 @@ from tasks.sync_website_document_indexing_task import sync_website_document_inde
 
 class DatasetService:
     @staticmethod
-    def get_datasets(page, per_page, provider="vendor", tenant_id=None, user=None, search=None, tag_ids=None):
-        query = Dataset.query.filter(Dataset.provider == provider, Dataset.tenant_id == tenant_id).order_by(
-            Dataset.created_at.desc()
-        )
+    def get_datasets(page, per_page, tenant_id=None, user=None, search=None, tag_ids=None):
+        query = Dataset.query.filter(Dataset.tenant_id == tenant_id).order_by(Dataset.created_at.desc())
 
         if user:
             # get permitted dataset ids
@@ -137,7 +137,14 @@ class DatasetService:
 
     @staticmethod
     def create_empty_dataset(
-        tenant_id: str, name: str, indexing_technique: Optional[str], account: Account, permission: Optional[str] = None
+        tenant_id: str,
+        name: str,
+        indexing_technique: Optional[str],
+        account: Account,
+        permission: Optional[str] = None,
+        provider: str = "vendor",
+        external_knowledge_api_id: Optional[str] = None,
+        external_knowledge_id: Optional[str] = None,
     ):
         # check if dataset name already exists
         if Dataset.query.filter_by(name=name, tenant_id=tenant_id).first():
@@ -156,12 +163,28 @@ class DatasetService:
         dataset.embedding_model_provider = embedding_model.provider if embedding_model else None
         dataset.embedding_model = embedding_model.model if embedding_model else None
         dataset.permission = permission or DatasetPermissionEnum.ONLY_ME
+        dataset.provider = provider
         db.session.add(dataset)
+        db.session.flush()
+
+        if provider == "external" and external_knowledge_api_id:
+            external_knowledge_api = ExternalDatasetService.get_external_knowledge_api(external_knowledge_api_id)
+            if not external_knowledge_api:
+                raise ValueError("External API template not found.")
+            external_knowledge_binding = ExternalKnowledgeBindings(
+                tenant_id=tenant_id,
+                dataset_id=dataset.id,
+                external_knowledge_api_id=external_knowledge_api_id,
+                external_knowledge_id=external_knowledge_id,
+                created_by=account.id,
+            )
+            db.session.add(external_knowledge_binding)
+
         db.session.commit()
         return dataset
 
     @staticmethod
-    def get_dataset(dataset_id):
+    def get_dataset(dataset_id) -> Dataset:
         return Dataset.query.filter_by(id=dataset_id).first()
 
     @staticmethod
@@ -202,81 +225,103 @@ class DatasetService:
 
     @staticmethod
     def update_dataset(dataset_id, data, user):
-        data.pop("partial_member_list", None)
-        filtered_data = {k: v for k, v in data.items() if v is not None or k == "description"}
         dataset = DatasetService.get_dataset(dataset_id)
+
         DatasetService.check_dataset_permission(dataset, user)
-        action = None
-        if dataset.indexing_technique != data["indexing_technique"]:
-            # if update indexing_technique
-            if data["indexing_technique"] == "economy":
-                action = "remove"
-                filtered_data["embedding_model"] = None
-                filtered_data["embedding_model_provider"] = None
-                filtered_data["collection_binding_id"] = None
-            elif data["indexing_technique"] == "high_quality":
-                action = "add"
-                # get embedding model setting
-                try:
-                    model_manager = ModelManager()
-                    embedding_model = model_manager.get_model_instance(
-                        tenant_id=current_user.current_tenant_id,
-                        provider=data["embedding_model_provider"],
-                        model_type=ModelType.TEXT_EMBEDDING,
-                        model=data["embedding_model"],
-                    )
-                    filtered_data["embedding_model"] = embedding_model.model
-                    filtered_data["embedding_model_provider"] = embedding_model.provider
-                    dataset_collection_binding = DatasetCollectionBindingService.get_dataset_collection_binding(
-                        embedding_model.provider, embedding_model.model
-                    )
-                    filtered_data["collection_binding_id"] = dataset_collection_binding.id
-                except LLMBadRequestError:
-                    raise ValueError(
-                        "No Embedding Model available. Please configure a valid provider "
-                        "in the Settings -> Model Provider."
-                    )
-                except ProviderTokenNotInitError as ex:
-                    raise ValueError(ex.description)
-        else:
+        if dataset.provider == "external":
+            dataset.retrieval_model = data.get("external_retrieval_model", None)
+            dataset.name = data.get("name", dataset.name)
+            dataset.description = data.get("description", "")
+            external_knowledge_id = data.get("external_knowledge_id", None)
+            db.session.add(dataset)
+            if not external_knowledge_id:
+                raise ValueError("External knowledge id is required.")
+            external_knowledge_api_id = data.get("external_knowledge_api_id", None)
+            if not external_knowledge_api_id:
+                raise ValueError("External knowledge api id is required.")
+            external_knowledge_binding = ExternalKnowledgeBindings.query.filter_by(dataset_id=dataset_id).first()
             if (
-                data["embedding_model_provider"] != dataset.embedding_model_provider
-                or data["embedding_model"] != dataset.embedding_model
+                external_knowledge_binding.external_knowledge_id != external_knowledge_id
+                or external_knowledge_binding.external_knowledge_api_id != external_knowledge_api_id
             ):
-                action = "update"
-                try:
-                    model_manager = ModelManager()
-                    embedding_model = model_manager.get_model_instance(
-                        tenant_id=current_user.current_tenant_id,
-                        provider=data["embedding_model_provider"],
-                        model_type=ModelType.TEXT_EMBEDDING,
-                        model=data["embedding_model"],
-                    )
-                    filtered_data["embedding_model"] = embedding_model.model
-                    filtered_data["embedding_model_provider"] = embedding_model.provider
-                    dataset_collection_binding = DatasetCollectionBindingService.get_dataset_collection_binding(
-                        embedding_model.provider, embedding_model.model
-                    )
-                    filtered_data["collection_binding_id"] = dataset_collection_binding.id
-                except LLMBadRequestError:
-                    raise ValueError(
-                        "No Embedding Model available. Please configure a valid provider "
-                        "in the Settings -> Model Provider."
-                    )
-                except ProviderTokenNotInitError as ex:
-                    raise ValueError(ex.description)
+                external_knowledge_binding.external_knowledge_id = external_knowledge_id
+                external_knowledge_binding.external_knowledge_api_id = external_knowledge_api_id
+                db.session.add(external_knowledge_binding)
+            db.session.commit()
+        else:
+            data.pop("partial_member_list", None)
+            filtered_data = {k: v for k, v in data.items() if v is not None or k == "description"}
+            action = None
+            if dataset.indexing_technique != data["indexing_technique"]:
+                # if update indexing_technique
+                if data["indexing_technique"] == "economy":
+                    action = "remove"
+                    filtered_data["embedding_model"] = None
+                    filtered_data["embedding_model_provider"] = None
+                    filtered_data["collection_binding_id"] = None
+                elif data["indexing_technique"] == "high_quality":
+                    action = "add"
+                    # get embedding model setting
+                    try:
+                        model_manager = ModelManager()
+                        embedding_model = model_manager.get_model_instance(
+                            tenant_id=current_user.current_tenant_id,
+                            provider=data["embedding_model_provider"],
+                            model_type=ModelType.TEXT_EMBEDDING,
+                            model=data["embedding_model"],
+                        )
+                        filtered_data["embedding_model"] = embedding_model.model
+                        filtered_data["embedding_model_provider"] = embedding_model.provider
+                        dataset_collection_binding = DatasetCollectionBindingService.get_dataset_collection_binding(
+                            embedding_model.provider, embedding_model.model
+                        )
+                        filtered_data["collection_binding_id"] = dataset_collection_binding.id
+                    except LLMBadRequestError:
+                        raise ValueError(
+                            "No Embedding Model available. Please configure a valid provider "
+                            "in the Settings -> Model Provider."
+                        )
+                    except ProviderTokenNotInitError as ex:
+                        raise ValueError(ex.description)
+            else:
+                if (
+                    data["embedding_model_provider"] != dataset.embedding_model_provider
+                    or data["embedding_model"] != dataset.embedding_model
+                ):
+                    action = "update"
+                    try:
+                        model_manager = ModelManager()
+                        embedding_model = model_manager.get_model_instance(
+                            tenant_id=current_user.current_tenant_id,
+                            provider=data["embedding_model_provider"],
+                            model_type=ModelType.TEXT_EMBEDDING,
+                            model=data["embedding_model"],
+                        )
+                        filtered_data["embedding_model"] = embedding_model.model
+                        filtered_data["embedding_model_provider"] = embedding_model.provider
+                        dataset_collection_binding = DatasetCollectionBindingService.get_dataset_collection_binding(
+                            embedding_model.provider, embedding_model.model
+                        )
+                        filtered_data["collection_binding_id"] = dataset_collection_binding.id
+                    except LLMBadRequestError:
+                        raise ValueError(
+                            "No Embedding Model available. Please configure a valid provider "
+                            "in the Settings -> Model Provider."
+                        )
+                    except ProviderTokenNotInitError as ex:
+                        raise ValueError(ex.description)
 
-        filtered_data["updated_by"] = user.id
-        filtered_data["updated_at"] = datetime.datetime.now()
+            filtered_data["updated_by"] = user.id
+            filtered_data["updated_at"] = datetime.datetime.now()
 
-        # update Retrieval model
-        filtered_data["retrieval_model"] = data["retrieval_model"]
+            # update Retrieval model
+            filtered_data["retrieval_model"] = data["retrieval_model"]
 
-        dataset.query.filter_by(id=dataset_id).update(filtered_data)
+            dataset.query.filter_by(id=dataset_id).update(filtered_data)
 
-        db.session.commit()
-        if action:
-            deal_dataset_vector_index_task.delay(dataset_id, action)
+            db.session.commit()
+            if action:
+                deal_dataset_vector_index_task.delay(dataset_id, action)
         return dataset
 
     @staticmethod

api/services/entities/external_knowledge_entities/external_knowledge_entities.py

@@ -0,0 +1,26 @@
+from typing import Literal, Optional, Union
+
+from pydantic import BaseModel
+
+
+class AuthorizationConfig(BaseModel):
+    type: Literal[None, "basic", "bearer", "custom"]
+    api_key: Union[None, str] = None
+    header: Union[None, str] = None
+
+
+class Authorization(BaseModel):
+    type: Literal["no-auth", "api-key"]
+    config: Optional[AuthorizationConfig] = None
+
+
+class ProcessStatusSetting(BaseModel):
+    request_method: str
+    url: str
+
+
+class ExternalKnowledgeApiSetting(BaseModel):
+    url: str
+    request_method: str
+    headers: Optional[dict] = None
+    params: Optional[dict] = None

api/services/external_knowledge_service.py

@@ -0,0 +1,378 @@
+import json
+import random
+import time
+from copy import deepcopy
+from datetime import datetime, timezone
+from typing import Any, Optional, Union
+
+import boto3
+import httpx
+import validators
+
+# from tasks.external_document_indexing_task import external_document_indexing_task
+from configs import dify_config
+from core.helper import ssrf_proxy
+from extensions.ext_database import db
+from models.dataset import (
+    Dataset,
+    Document,
+    ExternalKnowledgeApis,
+    ExternalKnowledgeBindings,
+)
+from models.model import UploadFile
+from services.entities.external_knowledge_entities.external_knowledge_entities import (
+    Authorization,
+    ExternalKnowledgeApiSetting,
+)
+from services.errors.dataset import DatasetNameDuplicateError
+
+
+class ExternalDatasetService:
+    @staticmethod
+    def get_external_knowledge_apis(page, per_page, tenant_id, search=None) -> tuple[list[ExternalKnowledgeApis], int]:
+        query = ExternalKnowledgeApis.query.filter(ExternalKnowledgeApis.tenant_id == tenant_id).order_by(
+            ExternalKnowledgeApis.created_at.desc()
+        )
+        if search:
+            query = query.filter(ExternalKnowledgeApis.name.ilike(f"%{search}%"))
+
+        external_knowledge_apis = query.paginate(page=page, per_page=per_page, max_per_page=100, error_out=False)
+
+        return external_knowledge_apis.items, external_knowledge_apis.total
+
+    @classmethod
+    def validate_api_list(cls, api_settings: dict):
+        if not api_settings:
+            raise ValueError("api list is empty")
+        if "endpoint" not in api_settings and not api_settings["endpoint"]:
+            raise ValueError("endpoint is required")
+        if "api_key" not in api_settings and not api_settings["api_key"]:
+            raise ValueError("api_key is required")
+
+    @staticmethod
+    def create_external_knowledge_api(tenant_id: str, user_id: str, args: dict) -> ExternalKnowledgeApis:
+        ExternalDatasetService.check_endpoint_and_api_key(args.get("settings"))
+        external_knowledge_api = ExternalKnowledgeApis(
+            tenant_id=tenant_id,
+            created_by=user_id,
+            updated_by=user_id,
+            name=args.get("name"),
+            description=args.get("description", ""),
+            settings=json.dumps(args.get("settings"), ensure_ascii=False),
+        )
+
+        db.session.add(external_knowledge_api)
+        db.session.commit()
+        return external_knowledge_api
+
+    @staticmethod
+    def check_endpoint_and_api_key(settings: dict):
+        if "endpoint" not in settings or not settings["endpoint"]:
+            raise ValueError("endpoint is required")
+        if "api_key" not in settings or not settings["api_key"]:
+            raise ValueError("api_key is required")
+
+        endpoint = f"{settings['endpoint']}/retrieval"
+        api_key = settings["api_key"]
+        if not validators.url(endpoint):
+            raise ValueError(f"invalid endpoint: {endpoint}")
+        try:
+            response = httpx.post(endpoint, headers={"Authorization": f"Bearer {api_key}"})
+        except Exception as e:
+            raise ValueError(f"failed to connect to the endpoint: {endpoint}")
+        if response.status_code == 502:
+            raise ValueError(f"Bad Gateway: failed to connect to the endpoint: {endpoint}")
+        if response.status_code == 404:
+            raise ValueError(f"Not Found: failed to connect to the endpoint: {endpoint}")
+        if response.status_code == 403:
+            raise ValueError(f"Forbidden: Authorization failed with api_key: {api_key}")
+
+    @staticmethod
+    def get_external_knowledge_api(external_knowledge_api_id: str) -> ExternalKnowledgeApis:
+        return ExternalKnowledgeApis.query.filter_by(id=external_knowledge_api_id).first()
+
+    @staticmethod
+    def update_external_knowledge_api(tenant_id, user_id, external_knowledge_api_id, args) -> ExternalKnowledgeApis:
+        external_knowledge_api = ExternalKnowledgeApis.query.filter_by(
+            id=external_knowledge_api_id, tenant_id=tenant_id
+        ).first()
+        if external_knowledge_api is None:
+            raise ValueError("api template not found")
+
+        external_knowledge_api.name = args.get("name")
+        external_knowledge_api.description = args.get("description", "")
+        external_knowledge_api.settings = json.dumps(args.get("settings"), ensure_ascii=False)
+        external_knowledge_api.updated_by = user_id
+        external_knowledge_api.updated_at = datetime.now(timezone.utc).replace(tzinfo=None)
+        db.session.commit()
+
+        return external_knowledge_api
+
+    @staticmethod
+    def delete_external_knowledge_api(tenant_id: str, external_knowledge_api_id: str):
+        external_knowledge_api = ExternalKnowledgeApis.query.filter_by(
+            id=external_knowledge_api_id, tenant_id=tenant_id
+        ).first()
+        if external_knowledge_api is None:
+            raise ValueError("api template not found")
+
+        db.session.delete(external_knowledge_api)
+        db.session.commit()
+
+    @staticmethod
+    def external_knowledge_api_use_check(external_knowledge_api_id: str) -> tuple[bool, int]:
+        count = ExternalKnowledgeBindings.query.filter_by(external_knowledge_api_id=external_knowledge_api_id).count()
+        if count > 0:
+            return True, count
+        return False, 0
+
+    @staticmethod
+    def get_external_knowledge_binding_with_dataset_id(tenant_id: str, dataset_id: str) -> ExternalKnowledgeBindings:
+        external_knowledge_binding = ExternalKnowledgeBindings.query.filter_by(
+            dataset_id=dataset_id, tenant_id=tenant_id
+        ).first()
+        if not external_knowledge_binding:
+            raise ValueError("external knowledge binding not found")
+        return external_knowledge_binding
+
+    @staticmethod
+    def document_create_args_validate(tenant_id: str, external_knowledge_api_id: str, process_parameter: dict):
+        external_knowledge_api = ExternalKnowledgeApis.query.filter_by(
+            id=external_knowledge_api_id, tenant_id=tenant_id
+        ).first()
+        if external_knowledge_api is None:
+            raise ValueError("api template not found")
+        settings = json.loads(external_knowledge_api.settings)
+        for setting in settings:
+            custom_parameters = setting.get("document_process_setting")
+            if custom_parameters:
+                for parameter in custom_parameters:
+                    if parameter.get("required", False) and not process_parameter.get(parameter.get("name")):
+                        raise ValueError(f'{parameter.get("name")} is required')
+
+    @staticmethod
+    def init_external_dataset(tenant_id: str, user_id: str, args: dict, created_from: str = "web"):
+        external_knowledge_api_id = args.get("external_knowledge_api_id")
+
+        data_source = args.get("data_source")
+        if data_source is None:
+            raise ValueError("data source is required")
+
+        process_parameter = args.get("process_parameter")
+        external_knowledge_api = ExternalKnowledgeApis.query.filter_by(
+            id=external_knowledge_api_id, tenant_id=tenant_id
+        ).first()
+        if external_knowledge_api is None:
+            raise ValueError("api template not found")
+
+        dataset = Dataset(
+            tenant_id=tenant_id,
+            name=args.get("name"),
+            description=args.get("description", ""),
+            provider="external",
+            created_by=user_id,
+        )
+
+        db.session.add(dataset)
+        db.session.flush()
+
+        document = Document.query.filter_by(dataset_id=dataset.id).order_by(Document.position.desc()).first()
+
+        position = document.position + 1 if document else 1
+
+        batch = time.strftime("%Y%m%d%H%M%S") + str(random.randint(100000, 999999))
+        document_ids = []
+        if data_source["type"] == "upload_file":
+            upload_file_list = data_source["info_list"]["file_info_list"]["file_ids"]
+            for file_id in upload_file_list:
+                file = (
+                    db.session.query(UploadFile)
+                    .filter(UploadFile.tenant_id == dataset.tenant_id, UploadFile.id == file_id)
+                    .first()
+                )
+                if file:
+                    data_source_info = {
+                        "upload_file_id": file_id,
+                    }
+                    document = Document(
+                        tenant_id=dataset.tenant_id,
+                        dataset_id=dataset.id,
+                        position=position,
+                        data_source_type=data_source["type"],
+                        data_source_info=json.dumps(data_source_info),
+                        batch=batch,
+                        name=file.name,
+                        created_from=created_from,
+                        created_by=user_id,
+                    )
+                    position += 1
+                    db.session.add(document)
+                    db.session.flush()
+                    document_ids.append(document.id)
+        db.session.commit()
+        # external_document_indexing_task.delay(dataset.id, external_knowledge_api_id, data_source, process_parameter)
+
+        return dataset
+
+    @staticmethod
+    def process_external_api(
+        settings: ExternalKnowledgeApiSetting, files: Union[None, dict[str, Any]]
+    ) -> httpx.Response:
+        """
+        do http request depending on api bundle
+        """
+
+        kwargs = {
+            "url": settings.url,
+            "headers": settings.headers,
+            "follow_redirects": True,
+        }
+
+        response = getattr(ssrf_proxy, settings.request_method)(data=json.dumps(settings.params), files=files, **kwargs)
+
+        return response
+
+    @staticmethod
+    def assembling_headers(authorization: Authorization, headers: Optional[dict] = None) -> dict[str, Any]:
+        authorization = deepcopy(authorization)
+        if headers:
+            headers = deepcopy(headers)
+        else:
+            headers = {}
+        if authorization.type == "api-key":
+            if authorization.config is None:
+                raise ValueError("authorization config is required")
+
+            if authorization.config.api_key is None:
+                raise ValueError("api_key is required")
+
+            if not authorization.config.header:
+                authorization.config.header = "Authorization"
+
+            if authorization.config.type == "bearer":
+                headers[authorization.config.header] = f"Bearer {authorization.config.api_key}"
+            elif authorization.config.type == "basic":
+                headers[authorization.config.header] = f"Basic {authorization.config.api_key}"
+            elif authorization.config.type == "custom":
+                headers[authorization.config.header] = authorization.config.api_key
+
+        return headers
+
+    @staticmethod
+    def get_external_knowledge_api_settings(settings: dict) -> ExternalKnowledgeApiSetting:
+        return ExternalKnowledgeApiSetting.parse_obj(settings)
+
+    @staticmethod
+    def create_external_dataset(tenant_id: str, user_id: str, args: dict) -> Dataset:
+        # check if dataset name already exists
+        if Dataset.query.filter_by(name=args.get("name"), tenant_id=tenant_id).first():
+            raise DatasetNameDuplicateError(f"Dataset with name {args.get('name')} already exists.")
+        external_knowledge_api = ExternalKnowledgeApis.query.filter_by(
+            id=args.get("external_knowledge_api_id"), tenant_id=tenant_id
+        ).first()
+
+        if external_knowledge_api is None:
+            raise ValueError("api template not found")
+
+        dataset = Dataset(
+            tenant_id=tenant_id,
+            name=args.get("name"),
+            description=args.get("description", ""),
+            provider="external",
+            retrieval_model=args.get("external_retrieval_model"),
+            created_by=user_id,
+        )
+
+        db.session.add(dataset)
+        db.session.flush()
+
+        external_knowledge_binding = ExternalKnowledgeBindings(
+            tenant_id=tenant_id,
+            dataset_id=dataset.id,
+            external_knowledge_api_id=args.get("external_knowledge_api_id"),
+            external_knowledge_id=args.get("external_knowledge_id"),
+            created_by=user_id,
+        )
+        db.session.add(external_knowledge_binding)
+
+        db.session.commit()
+
+        return dataset
+
+    @staticmethod
+    def fetch_external_knowledge_retrieval(
+        tenant_id: str, dataset_id: str, query: str, external_retrieval_parameters: dict
+    ) -> list:
+        external_knowledge_binding = ExternalKnowledgeBindings.query.filter_by(
+            dataset_id=dataset_id, tenant_id=tenant_id
+        ).first()
+        if not external_knowledge_binding:
+            raise ValueError("external knowledge binding not found")
+
+        external_knowledge_api = ExternalKnowledgeApis.query.filter_by(
+            id=external_knowledge_binding.external_knowledge_api_id
+        ).first()
+        if not external_knowledge_api:
+            raise ValueError("external api template not found")
+
+        settings = json.loads(external_knowledge_api.settings)
+        headers = {"Content-Type": "application/json"}
+        if settings.get("api_key"):
+            headers["Authorization"] = f"Bearer {settings.get('api_key')}"
+        score_threshold_enabled = external_retrieval_parameters.get("score_threshold_enabled") or False
+        score_threshold = external_retrieval_parameters.get("score_threshold", 0.0) if score_threshold_enabled else 0.0
+        request_params = {
+            "retrieval_setting": {
+                "top_k": external_retrieval_parameters.get("top_k"),
+                "score_threshold": score_threshold,
+            },
+            "query": query,
+            "knowledge_id": external_knowledge_binding.external_knowledge_id,
+        }
+
+        external_knowledge_api_setting = {
+            "url": f"{settings.get('endpoint')}/retrieval",
+            "request_method": "post",
+            "headers": headers,
+            "params": request_params,
+        }
+        response = ExternalDatasetService.process_external_api(
+            ExternalKnowledgeApiSetting(**external_knowledge_api_setting), None
+        )
+        if response.status_code == 200:
+            return response.json().get("records", [])
+        return []
+
+    @staticmethod
+    def test_external_knowledge_retrieval(retrieval_setting: dict, query: str, external_knowledge_id: str):
+        client = boto3.client(
+            "bedrock-agent-runtime",
+            aws_secret_access_key=dify_config.AWS_SECRET_ACCESS_KEY,
+            aws_access_key_id=dify_config.AWS_ACCESS_KEY_ID,
+            region_name="us-east-1",
+        )
+        response = client.retrieve(
+            knowledgeBaseId=external_knowledge_id,
+            retrievalConfiguration={
+                "vectorSearchConfiguration": {
+                    "numberOfResults": retrieval_setting.get("top_k"),
+                    "overrideSearchType": "HYBRID",
+                }
+            },
+            retrievalQuery={"text": query},
+        )
+        results = []
+        if response.get("ResponseMetadata") and response.get("ResponseMetadata").get("HTTPStatusCode") == 200:
+            if response.get("retrievalResults"):
+                retrieval_results = response.get("retrievalResults")
+                for retrieval_result in retrieval_results:
+                    if retrieval_result.get("score") < retrieval_setting.get("score_threshold", 0.0):
+                        continue
+                    result = {
+                        "metadata": retrieval_result.get("metadata"),
+                        "score": retrieval_result.get("score"),
+                        "title": retrieval_result.get("metadata").get("x-amz-bedrock-kb-source-uri"),
+                        "content": retrieval_result.get("content").get("text"),
+                    }
+                    results.append(result)
+        return {"records": results}

api/services/hit_testing_service.py

@@ -19,7 +19,15 @@ default_retrieval_model = {
 
 class HitTestingService:
     @classmethod
-    def retrieve(cls, dataset: Dataset, query: str, account: Account, retrieval_model: dict, limit: int = 10) -> dict:
+    def retrieve(
+        cls,
+        dataset: Dataset,
+        query: str,
+        account: Account,
+        retrieval_model: dict,
+        external_retrieval_model: dict,
+        limit: int = 10,
+    ) -> dict:
         if dataset.available_document_count == 0 or dataset.available_segment_count == 0:
             return {
                 "query": {
@@ -62,10 +70,44 @@ class HitTestingService:
 
         return cls.compact_retrieve_response(dataset, query, all_documents)
 
+    @classmethod
+    def external_retrieve(
+        cls,
+        dataset: Dataset,
+        query: str,
+        account: Account,
+        external_retrieval_model: dict,
+    ) -> dict:
+        if dataset.provider != "external":
+            return {
+                "query": {"content": query},
+                "records": [],
+            }
+
+        start = time.perf_counter()
+
+        all_documents = RetrievalService.external_retrieve(
+            dataset_id=dataset.id,
+            query=cls.escape_query_for_search(query),
+            external_retrieval_model=external_retrieval_model,
+        )
+
+        end = time.perf_counter()
+        logging.debug(f"External knowledge hit testing retrieve in {end - start:0.4f} seconds")
+
+        dataset_query = DatasetQuery(
+            dataset_id=dataset.id, content=query, source="hit_testing", created_by_role="account", created_by=account.id
+        )
+
+        db.session.add(dataset_query)
+        db.session.commit()
+
+        return cls.compact_external_retrieve_response(dataset, query, all_documents)
+
     @classmethod
     def compact_retrieve_response(cls, dataset: Dataset, query: str, documents: list[Document]):
-        i = 0
         records = []
+
         for document in documents:
             index_node_id = document.metadata["doc_id"]
 
@@ -81,7 +123,6 @@ class HitTestingService:
             )
 
             if not segment:
-                i += 1
                 continue
 
             record = {
@@ -91,8 +132,6 @@ class HitTestingService:
 
             records.append(record)
 
-            i += 1
-
         return {
             "query": {
                 "content": query,
@@ -100,6 +139,25 @@ class HitTestingService:
             "records": records,
         }
 
+    @classmethod
+    def compact_external_retrieve_response(cls, dataset: Dataset, query: str, documents: list):
+        records = []
+        if dataset.provider == "external":
+            for document in documents:
+                record = {
+                    "content": document.get("content", None),
+                    "title": document.get("title", None),
+                    "score": document.get("score", None),
+                    "metadata": document.get("metadata", None),
+                }
+                records.append(record)
+            return {
+                "query": {
+                    "content": query,
+                },
+                "records": records,
+            }
+
     @classmethod
     def hit_testing_args_check(cls, args):
         query = args["query"]

api/tasks/external_document_indexing_task.py

@@ -0,0 +1,93 @@
+import json
+import logging
+import time
+
+import click
+from celery import shared_task
+
+from core.indexing_runner import DocumentIsPausedException
+from extensions.ext_database import db
+from extensions.ext_storage import storage
+from models.dataset import Dataset, ExternalKnowledgeApis
+from models.model import UploadFile
+from services.external_knowledge_service import ExternalDatasetService
+
+
+@shared_task(queue="dataset")
+def external_document_indexing_task(
+    dataset_id: str, external_knowledge_api_id: str, data_source: dict, process_parameter: dict
+):
+    """
+    Async process document
+    :param dataset_id:
+    :param external_knowledge_api_id:
+    :param data_source:
+    :param process_parameter:
+    Usage: external_document_indexing_task.delay(dataset_id, document_id)
+    """
+    start_at = time.perf_counter()
+
+    dataset = db.session.query(Dataset).filter(Dataset.id == dataset_id).first()
+    if not dataset:
+        logging.info(
+            click.style("Processed external dataset: {} failed, dataset not exit.".format(dataset_id), fg="red")
+        )
+        return
+
+    # get external api template
+    external_knowledge_api = (
+        db.session.query(ExternalKnowledgeApis)
+        .filter(
+            ExternalKnowledgeApis.id == external_knowledge_api_id, ExternalKnowledgeApis.tenant_id == dataset.tenant_id
+        )
+        .first()
+    )
+
+    if not external_knowledge_api:
+        logging.info(
+            click.style(
+                "Processed external dataset: {} failed, api template: {} not exit.".format(
+                    dataset_id, external_knowledge_api_id
+                ),
+                fg="red",
+            )
+        )
+        return
+    files = {}
+    if data_source["type"] == "upload_file":
+        upload_file_list = data_source["info_list"]["file_info_list"]["file_ids"]
+        for file_id in upload_file_list:
+            file = (
+                db.session.query(UploadFile)
+                .filter(UploadFile.tenant_id == dataset.tenant_id, UploadFile.id == file_id)
+                .first()
+            )
+            if file:
+                files[file.id] = (file.name, storage.load_once(file.key), file.mime_type)
+    try:
+        settings = ExternalDatasetService.get_external_knowledge_api_settings(
+            json.loads(external_knowledge_api.settings)
+        )
+        # assemble headers
+        headers = ExternalDatasetService.assembling_headers(settings.authorization, settings.headers)
+
+        # do http request
+        response = ExternalDatasetService.process_external_api(settings, headers, process_parameter, files)
+        job_id = response.json().get("job_id")
+        if job_id:
+            # save job_id to dataset
+            dataset.job_id = job_id
+            db.session.commit()
+
+        end_at = time.perf_counter()
+        logging.info(
+            click.style(
+                "Processed external dataset: {} successful, latency: {}".format(dataset.id, end_at - start_at),
+                fg="green",
+            )
+        )
+    except DocumentIsPausedException as ex:
+        logging.info(click.style(str(ex), fg="yellow"))
+
+    except Exception:
+        pass

docker/unstructured.yaml

@@ -0,0 +1,24 @@
+
+  # unstructured .
+  # (if used, you need to set ETL_TYPE to Unstructured in the api & worker service.)
+  unstructured:
+    image: downloads.unstructured.io/unstructured-io/unstructured-api:latest
+    profiles:
+      - unstructured
+    restart: always
+    volumes:
+      - ./volumes/unstructured:/app/data
+
+networks:
+  # create a network between sandbox, api and ssrf_proxy, and can not access outside.
+  ssrf_proxy_network:
+    driver: bridge
+    internal: true
+  milvus:
+    driver: bridge
+  opensearch-net:
+    driver: bridge
+    internal: true
+
+volumes:
+  oradata:

web/app/(commonLayout)/datasets/(datasetDetailLayout)/[datasetId]/layout.tsx

@@ -1,6 +1,6 @@
 'use client'
 import type { FC, SVGProps } from 'react'
-import React, { useEffect } from 'react'
+import React, { useEffect, useMemo } from 'react'
 import { usePathname } from 'next/navigation'
 import useSWR from 'swr'
 import { useTranslation } from 'react-i18next'
@@ -203,12 +203,23 @@ const DatasetDetailLayout: FC<IAppDetailLayoutProps> = (props) => {
     datasetId,
   }, apiParams => fetchDatasetRelatedApps(apiParams.datasetId))
 
-  const navigation = [
-    { name: t('common.datasetMenus.documents'), href: `/datasets/${datasetId}/documents`, icon: DocumentTextIcon, selectedIcon: DocumentTextSolidIcon },
-    { name: t('common.datasetMenus.hitTesting'), href: `/datasets/${datasetId}/hitTesting`, icon: TargetIcon, selectedIcon: TargetSolidIcon },
-    // { name: 'api & webhook', href: `/datasets/${datasetId}/api`, icon: CommandLineIcon, selectedIcon: CommandLineSolidIcon },
-    { name: t('common.datasetMenus.settings'), href: `/datasets/${datasetId}/settings`, icon: Cog8ToothIcon, selectedIcon: Cog8ToothSolidIcon },
-  ]
+  const navigation = useMemo(() => {
+    const baseNavigation = [
+      { name: t('common.datasetMenus.hitTesting'), href: `/datasets/${datasetId}/hitTesting`, icon: TargetIcon, selectedIcon: TargetSolidIcon },
+      // { name: 'api & webhook', href: `/datasets/${datasetId}/api`, icon: CommandLineIcon, selectedIcon: CommandLineSolidIcon },
+      { name: t('common.datasetMenus.settings'), href: `/datasets/${datasetId}/settings`, icon: Cog8ToothIcon, selectedIcon: Cog8ToothSolidIcon },
+    ]
+
+    if (datasetRes?.provider !== 'external') {
+      baseNavigation.unshift({
+        name: t('common.datasetMenus.documents'),
+        href: `/datasets/${datasetId}/documents`,
+        icon: DocumentTextIcon,
+        selectedIcon: DocumentTextSolidIcon,
+      })
+    }
+    return baseNavigation
+  }, [datasetRes?.provider, datasetId, t])
 
   useEffect(() => {
     if (datasetRes)
@@ -233,6 +244,7 @@ const DatasetDetailLayout: FC<IAppDetailLayoutProps> = (props) => {
         icon={datasetRes?.icon || 'https://static.dify.ai/images/dataset-default-icon.png'}
         icon_background={datasetRes?.icon_background || '#F5F5F5'}
         desc={datasetRes?.description || '--'}
+        isExternal={datasetRes?.provider === 'external'}
         navigation={navigation}
         extraInfo={!isCurrentWorkspaceDatasetOperator ? mode => <ExtraInfo isMobile={mode === 'collapse'} relatedApps={relatedApps} /> : undefined}
         iconType={datasetRes?.data_source_type === DataSourceType.NOTION ? 'notion' : 'dataset'}

web/app/(commonLayout)/datasets/Container.tsx

@@ -8,6 +8,7 @@ import { useDebounceFn } from 'ahooks'
 import useSWR from 'swr'
 
 // Components
+import ExternalAPIPanel from '../../components/datasets/external-api/external-api-panel'
 import Datasets from './Datasets'
 import DatasetFooter from './DatasetFooter'
 import ApiServer from './ApiServer'
@@ -16,6 +17,8 @@ import TabSliderNew from '@/app/components/base/tab-slider-new'
 import SearchInput from '@/app/components/base/search-input'
 import TagManagementModal from '@/app/components/base/tag-management'
 import TagFilter from '@/app/components/base/tag-management/filter'
+import Button from '@/app/components/base/button'
+import { ApiConnectionMod } from '@/app/components/base/icons/src/vender/solid/development'
 
 // Services
 import { fetchDatasetApiBaseUrl } from '@/service/datasets'
@@ -24,12 +27,14 @@ import { fetchDatasetApiBaseUrl } from '@/service/datasets'
 import { useTabSearchParams } from '@/hooks/use-tab-searchparams'
 import { useStore as useTagStore } from '@/app/components/base/tag-management/store'
 import { useAppContext } from '@/context/app-context'
+import { useExternalApiPanel } from '@/context/external-api-panel-context'
 
 const Container = () => {
   const { t } = useTranslation()
   const router = useRouter()
   const { currentWorkspace } = useAppContext()
   const showTagManagementModal = useTagStore(s => s.showTagManagementModal)
+  const { showExternalApiPanel, setShowExternalApiPanel } = useExternalApiPanel()
 
   const options = useMemo(() => {
     return [
@@ -66,7 +71,7 @@ const Container = () => {
   useEffect(() => {
     if (currentWorkspace.role === 'normal')
       return router.replace('/apps')
-  }, [currentWorkspace])
+  }, [currentWorkspace, router])
 
   return (
     <div ref={containerRef} className='grow relative flex flex-col bg-gray-100 overflow-y-auto'>
@@ -80,11 +85,18 @@ const Container = () => {
           <div className='flex items-center gap-2'>
             <TagFilter type='knowledge' value={tagFilterValue} onChange={handleTagsChange} />
             <SearchInput className='w-[200px]' value={keywords} onChange={handleKeywordsChange} />
+            <div className="w-[1px] h-4 bg-divider-regular" />
+            <Button
+              className='gap-0.5 shadows-shadow-xs'
+              onClick={() => setShowExternalApiPanel(true)}
+            >
+              <ApiConnectionMod className='w-4 h-4 text-components-button-secondary-text' />
+              <div className='flex px-0.5 justify-center items-center gap-1 text-components-button-secondary-text system-sm-medium'>{t('dataset.externalAPIPanelTitle')}</div>
+            </Button>
           </div>
         )}
         {activeTab === 'api' && data && <ApiServer apiBaseUrl={data.api_base_url || ''} />}
       </div>
-
       {activeTab === 'dataset' && (
         <>
           <Datasets containerRef={containerRef} tags={tagIDs} keywords={searchKeywords} />
@@ -94,10 +106,10 @@ const Container = () => {
           )}
         </>
       )}
-
       {activeTab === 'api' && data && <Doc apiBaseUrl={data.api_base_url || ''} />}
-    </div>
 
+      {showExternalApiPanel && <ExternalAPIPanel onClose={() => setShowExternalApiPanel(false)} />}
+    </div>
   )
 }
 

web/app/(commonLayout)/datasets/DatasetCard.tsx

@@ -18,6 +18,7 @@ import Divider from '@/app/components/base/divider'
 import RenameDatasetModal from '@/app/components/datasets/rename-modal'
 import type { Tag } from '@/app/components/base/tag-management/constant'
 import TagSelector from '@/app/components/base/tag-management/selector'
+import CornerLabel from '@/app/components/base/corner-label'
 import { useAppContext } from '@/context/app-context'
 
 export type DatasetCardProps = {
@@ -32,6 +33,7 @@ const DatasetCard = ({
   const { t } = useTranslation()
   const { notify } = useContext(ToastContext)
   const { push } = useRouter()
+  const EXTERNAL_PROVIDER = 'external' as const
 
   const { isCurrentWorkspaceDatasetOperator } = useAppContext()
   const [tags, setTags] = useState<Tag[]>(dataset.tags)
@@ -39,6 +41,7 @@ const DatasetCard = ({
   const [showRenameModal, setShowRenameModal] = useState(false)
   const [showConfirmDelete, setShowConfirmDelete] = useState(false)
   const [confirmMessage, setConfirmMessage] = useState<string>('')
+  const isExternalProvider = (provider: string): boolean => provider === EXTERNAL_PROVIDER
   const detectIsUsedByApp = useCallback(async () => {
     try {
       const { is_using: isUsedByApp } = await checkIsUsedInApp(dataset.id)
@@ -108,13 +111,16 @@ const DatasetCard = ({
   return (
     <>
       <div
-        className='group col-span-1 bg-white border-2 border-solid border-transparent rounded-xl shadow-sm min-h-[160px] flex flex-col transition-all duration-200 ease-in-out cursor-pointer hover:shadow-lg'
+        className='group relative col-span-1 bg-white border-[0.5px] border-solid border-transparent rounded-xl shadow-sm min-h-[160px] flex flex-col transition-all duration-200 ease-in-out cursor-pointer hover:shadow-lg'
         data-disable-nprogress={true}
         onClick={(e) => {
           e.preventDefault()
-          push(`/datasets/${dataset.id}/documents`)
+          isExternalProvider(dataset.provider)
+            ? push(`/datasets/${dataset.id}/hitTesting`)
+            : push(`/datasets/${dataset.id}/documents`)
         }}
       >
+        {isExternalProvider(dataset.provider) && <CornerLabel label='External' className='absolute right-0' labelClassName='rounded-tr-xl' />}
         <div className='flex pt-[14px] px-[14px] pb-3 h-[66px] items-center gap-3 grow-0 shrink-0'>
           <div className={cn(
             'shrink-0 flex items-center justify-center p-2.5 bg-[#F5F8FF] rounded-md border-[0.5px] border-[#E0EAFF]',
@@ -136,13 +142,20 @@ const DatasetCard = ({
             <div className='flex items-center mt-[1px] text-xs leading-[18px] text-gray-500'>
               <div
                 className={cn('truncate', (!dataset.embedding_available || !dataset.document_count) && 'opacity-50')}
-                title={`${dataset.document_count}${t('dataset.documentCount')} · ${Math.round(dataset.word_count / 1000)}${t('dataset.wordCount')} · ${dataset.app_count}${t('dataset.appCount')}`}
+                title={dataset.provider === 'external' ? `${dataset.app_count}${t('dataset.appCount')}` : `${dataset.document_count}${t('dataset.documentCount')} · ${Math.round(dataset.word_count / 1000)}${t('dataset.wordCount')} · ${dataset.app_count}${t('dataset.appCount')}`}
               >
-                <span>{dataset.document_count}{t('dataset.documentCount')}</span>
-                <span className='shrink-0 mx-0.5 w-1 text-gray-400'>·</span>
-                <span>{Math.round(dataset.word_count / 1000)}{t('dataset.wordCount')}</span>
-                <span className='shrink-0 mx-0.5 w-1 text-gray-400'>·</span>
-                <span>{dataset.app_count}{t('dataset.appCount')}</span>
+                {dataset.provider === 'external'
+                  ? <>
+                    <span>{dataset.app_count}{t('dataset.appCount')}</span>
+                  </>
+                  : <>
+                    <span>{dataset.document_count}{t('dataset.documentCount')}</span>
+                    <span className='shrink-0 mx-0.5 w-1 text-gray-400'>·</span>
+                    <span>{Math.round(dataset.word_count / 1000)}{t('dataset.wordCount')}</span>
+                    <span className='shrink-0 mx-0.5 w-1 text-gray-400'>·</span>
+                    <span>{dataset.app_count}{t('dataset.appCount')}</span>
+                  </>
+                }
               </div>
             </div>
           </div>

web/app/(commonLayout)/datasets/NewDatasetCard.tsx

@@ -4,21 +4,32 @@ import { forwardRef } from 'react'
 import { useTranslation } from 'react-i18next'
 import {
   RiAddLine,
+  RiArrowRightLine,
 } from '@remixicon/react'
 
 const CreateAppCard = forwardRef<HTMLAnchorElement>((_, ref) => {
   const { t } = useTranslation()
 
   return (
-    <a ref={ref} className='group flex flex-col col-span-1 bg-gray-200 border-[0.5px] border-black/5 rounded-xl min-h-[160px] transition-all duration-200 ease-in-out cursor-pointer hover:bg-white hover:shadow-lg' href='/datasets/create'>
-      <div className='shrink-0 flex items-center p-4 pb-3'>
-        <div className='w-10 h-10 flex items-center justify-center border border-gray-200 bg-gray-100 rounded-lg'>
-          <RiAddLine className='w-4 h-4 text-gray-500'/>
+    <div className='flex flex-col bg-background-default-dimm border-[0.5px] border-components-panel-border rounded-xl
+      min-h-[160px] transition-all duration-200 ease-in-out'
+    >
+      <a ref={ref} className='group flex flex-grow items-start p-4 cursor-pointer' href='/datasets/create'>
+        <div className='flex items-center gap-3'>
+          <div className='w-10 h-10 p-2 flex items-center justify-center border border-dashed border-divider-regular rounded-lg
+            bg-background-default-lighter group-hover:border-solid group-hover:border-effects-highlight group-hover:bg-background-default-dodge'
+          >
+            <RiAddLine className='w-4 h-4 text-text-tertiary group-hover:text-text-accent'/>
+          </div>
+          <div className='system-md-semibold text-text-secondary group-hover:text-text-accent'>{t('dataset.createDataset')}</div>
         </div>
-        <div className='ml-3 text-sm font-semibold leading-5 text-gray-800 group-hover:text-primary-600'>{t('dataset.createDataset')}</div>
-      </div>
-      <div className='mb-1 px-4 text-xs leading-normal text-gray-500 line-clamp-4'>{t('dataset.createDatasetIntro')}</div>
-    </a>
+      </a>
+      <div className='p-4 pt-0 text-text-tertiary system-xs-regular'>{t('dataset.createDatasetIntro')}</div>
+      <a className='group flex p-4 items-center gap-1 border-t-[0.5px] border-divider-subtle rounded-b-xl cursor-pointer' href='/datasets/connect'>
+        <div className='system-xs-medium text-text-tertiary group-hover:text-text-accent'>{t('dataset.connectDataset')}</div>
+        <RiArrowRightLine className='w-3.5 h-3.5 text-text-tertiary group-hover:text-text-accent' />
+      </a>
+    </div>
   )
 })
 

web/app/(commonLayout)/datasets/connect/page.tsx

@@ -0,0 +1,8 @@
+import React from 'react'
+import ExternalKnowledgeBaseConnector from '@/app/components/datasets/external-knowledge-base/connector'
+
+const ExternalKnowledgeBaseCreation = () => {
+  return <ExternalKnowledgeBaseConnector />
+}
+
+export default ExternalKnowledgeBaseCreation

web/app/(commonLayout)/datasets/layout.tsx

@@ -0,0 +1,14 @@
+'use client'
+
+import { ExternalApiPanelProvider } from '@/context/external-api-panel-context'
+import { ExternalKnowledgeApiProvider } from '@/context/external-knowledge-api-context'
+
+export default function DatasetsLayout({ children }: { children: React.ReactNode }) {
+  return (
+    <ExternalKnowledgeApiProvider>
+      <ExternalApiPanelProvider>
+        {children}
+      </ExternalApiPanelProvider>
+    </ExternalKnowledgeApiProvider>
+  )
+}

web/app/(commonLayout)/datasets/page.tsx

@@ -1,9 +1,7 @@
 import Container from './Container'
 
 const AppList = async () => {
-  return (
-    <Container />
-  )
+  return <Container />
 }
 
 export const metadata = {

web/app/(commonLayout)/datasets/store.ts

@@ -0,0 +1,11 @@
+import { create } from 'zustand'
+
+type DatasetStore = {
+  showExternalApiPanel: boolean
+  setShowExternalApiPanel: (show: boolean) => void
+}
+
+export const useDatasetStore = create<DatasetStore>(set => ({
+  showExternalApiPanel: false,
+  setShowExternalApiPanel: show => set({ showExternalApiPanel: show }),
+}))

web/app/components/app-sidebar/basic.tsx

@@ -1,4 +1,5 @@
 import React from 'react'
+import { useTranslation } from 'react-i18next'
 import AppIcon from '../base/app-icon'
 import Tooltip from '@/app/components/base/tooltip'
 
@@ -6,6 +7,7 @@ export type IAppBasicProps = {
   iconType?: 'app' | 'api' | 'dataset' | 'webapp' | 'notion'
   icon?: string
   icon_background?: string | null
+  isExternal?: boolean
   name: string
   type: string | React.ReactNode
   hoverTip?: string
@@ -52,7 +54,9 @@ const ICON_MAP = {
   notion: <AppIcon innerIcon={NotionSvg} className='!border-[0.5px] !border-indigo-100 !bg-white' />,
 }
 
-export default function AppBasic({ icon, icon_background, name, type, hoverTip, textStyle, mode = 'expand', iconType = 'app' }: IAppBasicProps) {
+export default function AppBasic({ icon, icon_background, name, isExternal, type, hoverTip, textStyle, mode = 'expand', iconType = 'app' }: IAppBasicProps) {
+  const { t } = useTranslation()
+
   return (
     <div className="flex items-start p-1">
       {icon && icon_background && iconType === 'app' && (
@@ -83,6 +87,7 @@ export default function AppBasic({ icon, icon_background, name, type, hoverTip,
           }
         </div>
         <div className={`text-xs font-normal text-gray-500 group-hover:text-gray-700 break-all ${textStyle?.extra ?? ''}`}>{type}</div>
+        <div className='text-text-tertiary system-2xs-medium-uppercase'>{isExternal ? t('dataset.externalTag') : ''}</div>
       </div>}
     </div>
   )

web/app/components/app-sidebar/index.tsx

@@ -15,6 +15,7 @@ export type IAppDetailNavProps = {
   iconType?: 'app' | 'dataset' | 'notion'
   title: string
   desc: string
+  isExternal?: boolean
   icon: string
   icon_background: string
   navigation: Array<{
@@ -26,7 +27,7 @@ export type IAppDetailNavProps = {
   extraInfo?: (modeState: string) => React.ReactNode
 }
 
-const AppDetailNav = ({ title, desc, icon, icon_background, navigation, extraInfo, iconType = 'app' }: IAppDetailNavProps) => {
+const AppDetailNav = ({ title, desc, isExternal, icon, icon_background, navigation, extraInfo, iconType = 'app' }: IAppDetailNavProps) => {
   const { appSidebarExpand, setAppSiderbarExpand } = useAppStore(useShallow(state => ({
     appSidebarExpand: state.appSidebarExpand,
     setAppSiderbarExpand: state.setAppSiderbarExpand,
@@ -70,6 +71,7 @@ const AppDetailNav = ({ title, desc, icon, icon_background, navigation, extraInf
             icon_background={icon_background}
             name={title}
             type={desc}
+            isExternal={isExternal}
           />
         )}
       </div>

web/app/components/app/configuration/dataset-config/card-item/item.tsx

@@ -5,6 +5,7 @@ import {
   RiDeleteBinLine,
   RiEditLine,
 } from '@remixicon/react'
+import { useTranslation } from 'react-i18next'
 import SettingsModal from '../settings-modal'
 import type { DataSet } from '@/models/datasets'
 import { DataSourceType } from '@/models/datasets'
@@ -33,6 +34,7 @@ const Item: FC<ItemProps> = ({
   const isMobile = media === MediaType.mobile
   const [showSettingsModal, setShowSettingsModal] = useState(false)
   const { formatIndexingTechniqueAndMethod } = useKnowledge()
+  const { t } = useTranslation()
 
   const handleSave = (newDataset: DataSet) => {
     onSave(newDataset)
@@ -65,9 +67,11 @@ const Item: FC<ItemProps> = ({
       <div className='grow'>
         <div className='flex items-center h-[18px]'>
           <div className='grow text-[13px] font-medium text-gray-800 truncate' title={config.name}>{config.name}</div>
-          <Badge
-            text={formatIndexingTechniqueAndMethod(config.indexing_technique, config.retrieval_model_dict?.search_method)}
-          />
+          {config.provider === 'external'
+            ? <Badge text={t('dataset.externalTag')}></Badge>
+            : <Badge
+              text={formatIndexingTechniqueAndMethod(config.indexing_technique, config.retrieval_model_dict?.search_method)}
+            />}
         </div>
       </div>
       <div className='hidden rounded-lg group-hover:flex items-center justify-end absolute right-0 top-0 bottom-0 pr-2 w-[124px] bg-gradient-to-r from-white/50 to-white to-50%'>

web/app/components/app/configuration/dataset-config/params-config/config-content.tsx

@@ -174,6 +174,20 @@ const ConfigContent: FC<Props> = ({
               </div>
             )
           }
+          {
+            selectedDatasetsMode.mixtureInternalAndExternal && (
+              <div className='mt-4 system-xs-medium text-text-warning'>
+                {t('dataset.mixtureInternalAndExternalTip')}
+              </div>
+            )
+          }
+          {
+            selectedDatasetsMode.allExternal && (
+              <div className='mt-4 system-xs-medium text-text-warning'>
+                {t('dataset.allExternalTip')}
+              </div>
+            )
+          }
           {
             selectedDatasetsMode.mixtureHighQualityAndEconomic
             && (
@@ -229,15 +243,15 @@ const ConfigContent: FC<Props> = ({
                       />
                     )
                   }
-                  <div className='ml-2 leading-[32px] text-[13px] font-medium text-gray-900'>{t('common.modelProvider.rerankModel.key')}</div>
+                  <div className='leading-[32px] text-text-secondary system-sm-semibold ml-1'>{t('common.modelProvider.rerankModel.key')}</div>
                   <Tooltip
                     popupContent={
                       <div className="w-[200px]">
                         {t('common.modelProvider.rerankModel.tip')}
                       </div>
                     }
-                    popupClassName='ml-0.5'
-                    triggerClassName='ml-0.5 w-3.5 h-3.5'
+                    popupClassName='ml-1'
+                    triggerClassName='ml-1 w-4 h-4'
                   />
                 </div>
                 <div>

web/app/components/app/configuration/dataset-config/params-config/index.tsx

@@ -39,13 +39,26 @@ const ParamsConfig = ({
   useEffect(() => {
     const {
       allEconomic,
+      allHighQuality,
+      allHighQualityFullTextSearch,
+      allHighQualityVectorSearch,
+      allExternal,
+      mixtureHighQualityAndEconomic,
+      inconsistentEmbeddingModel,
+      mixtureInternalAndExternal,
     } = getSelectedDatasetsMode(selectedDatasets)
     const { datasets, retrieval_model, score_threshold_enabled, ...restConfigs } = datasetConfigs
     let rerankEnable = restConfigs.reranking_enable
 
-    if (allEconomic && !restConfigs.reranking_model?.reranking_provider_name && rerankEnable === undefined)
+    if ((allEconomic && !restConfigs.reranking_model?.reranking_provider_name && rerankEnable === undefined) || allExternal)
       rerankEnable = false
 
+    if (allEconomic || allHighQuality || allHighQualityFullTextSearch || allHighQualityVectorSearch || (allExternal && selectedDatasets.length === 1))
+      setRerankSettingModalOpen(false)
+
+    if (mixtureHighQualityAndEconomic || inconsistentEmbeddingModel || mixtureInternalAndExternal || (allExternal && selectedDatasets.length > 1))
+      setRerankSettingModalOpen(true)
+
     setTempDataSetConfigs({
       ...getMultipleRetrievalConfig({
         top_k: restConfigs.top_k,

web/app/components/app/configuration/dataset-config/select-dataset/index.tsx

@@ -47,7 +47,7 @@ const SelectDataSet: FC<ISelectDataSetProps> = ({
         const { data, has_more } = await fetchDatasets({ url: '/datasets', params: { page } })
         setPage(getPage() + 1)
         setIsNoMore(!has_more)
-        const newList = [...(datasets || []), ...data.filter(item => item.indexing_technique)]
+        const newList = [...(datasets || []), ...data.filter(item => item.indexing_technique || item.provider === 'external')]
         setDataSets(newList)
         setLoaded(true)
         if (!selected.find(item => !item.name))
@@ -145,6 +145,11 @@ const SelectDataSet: FC<ISelectDataSetProps> = ({
                     />
                   )
                 }
+                {
+                  item.provider === 'external' && (
+                    <Badge text={t('dataset.externalTag')} />
+                  )
+                }
               </div>
             ))}
           </div>

web/app/components/app/configuration/dataset-config/settings-modal/index.tsx

@@ -5,8 +5,10 @@ import { useTranslation } from 'react-i18next'
 import { isEqual } from 'lodash-es'
 import { RiCloseLine } from '@remixicon/react'
 import { BookOpenIcon } from '@heroicons/react/24/outline'
+import { ApiConnectionMod } from '@/app/components/base/icons/src/vender/solid/development'
 import cn from '@/utils/classnames'
 import IndexMethodRadio from '@/app/components/datasets/settings/index-method-radio'
+import Divider from '@/app/components/base/divider'
 import Button from '@/app/components/base/button'
 import type { DataSet } from '@/models/datasets'
 import { useToastContext } from '@/app/components/base/toast'
@@ -14,6 +16,7 @@ import { updateDatasetSetting } from '@/service/datasets'
 import { useAppContext } from '@/context/app-context'
 import { useModalContext } from '@/context/modal-context'
 import type { RetrievalConfig } from '@/types/app'
+import RetrievalSettings from '@/app/components/datasets/external-knowledge-base/create/RetrievalSettings'
 import RetrievalMethodConfig from '@/app/components/datasets/common/retrieval-method-config'
 import EconomicalRetrievalMethodConfig from '@/app/components/datasets/common/economical-retrieval-method-config'
 import { ensureRerankModelSelected, isReRankModelSelected } from '@/app/components/datasets/common/check-rerank-model'
@@ -56,6 +59,9 @@ const SettingsModal: FC<SettingsModalProps> = ({
   const { t } = useTranslation()
   const { notify } = useToastContext()
   const ref = useRef(null)
+  const [topK, setTopK] = useState(currentDataset?.external_retrieval_model.top_k ?? 2)
+  const [scoreThreshold, setScoreThreshold] = useState(currentDataset?.external_retrieval_model.score_threshold ?? 0.5)
+  const [scoreThresholdEnabled, setScoreThresholdEnabled] = useState(currentDataset?.external_retrieval_model.score_threshold_enabled ?? false)
 
   const { setShowAccountSettingModal } = useModalContext()
   const [loading, setLoading] = useState(false)
@@ -73,6 +79,15 @@ const SettingsModal: FC<SettingsModalProps> = ({
   const [isHideChangedTip, setIsHideChangedTip] = useState(false)
   const isRetrievalChanged = !isEqual(retrievalConfig, localeCurrentDataset?.retrieval_model_dict) || indexMethod !== localeCurrentDataset?.indexing_technique
 
+  const handleSettingsChange = (data: { top_k?: number; score_threshold?: number; score_threshold_enabled?: boolean }) => {
+    if (data.top_k !== undefined)
+      setTopK(data.top_k)
+    if (data.score_threshold !== undefined)
+      setScoreThreshold(data.score_threshold)
+    if (data.score_threshold_enabled !== undefined)
+      setScoreThresholdEnabled(data.score_threshold_enabled)
+  }
+
   const handleSave = async () => {
     if (loading)
       return
@@ -107,10 +122,17 @@ const SettingsModal: FC<SettingsModalProps> = ({
           description,
           permission,
           indexing_technique: indexMethod,
+          external_retrieval_model: {
+            top_k: topK,
+            score_threshold: scoreThreshold,
+            score_threshold_enabled: scoreThresholdEnabled,
+          },
           retrieval_model: {
             ...postRetrievalConfig,
             score_threshold: postRetrievalConfig.score_threshold_enabled ? postRetrievalConfig.score_threshold : 0,
           },
+          external_knowledge_id: currentDataset!.external_knowledge_info.external_knowledge_id,
+          external_knowledge_api_id: currentDataset!.external_knowledge_info.external_knowledge_api_id,
           embedding_model: localeCurrentDataset.embedding_model,
           embedding_model_provider: localeCurrentDataset.embedding_model_provider,
         },
@@ -178,7 +200,7 @@ const SettingsModal: FC<SettingsModalProps> = ({
       }}>
         <div className={cn(rowClass, 'items-center')}>
           <div className={labelClass}>
-            {t('datasetSettings.form.name')}
+            <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.name')}</div>
           </div>
           <input
             value={localeCurrentDataset.name}
@@ -189,7 +211,7 @@ const SettingsModal: FC<SettingsModalProps> = ({
         </div>
         <div className={cn(rowClass)}>
           <div className={labelClass}>
-            {t('datasetSettings.form.desc')}
+            <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.desc')}</div>
           </div>
           <div className='w-full'>
             <textarea
@@ -206,7 +228,7 @@ const SettingsModal: FC<SettingsModalProps> = ({
         </div>
         <div className={rowClass}>
           <div className={labelClass}>
-            <div>{t('datasetSettings.form.permissions')}</div>
+            <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.permissions')}</div>
           </div>
           <div className='w-full'>
             <PermissionSelector
@@ -219,24 +241,25 @@ const SettingsModal: FC<SettingsModalProps> = ({
             />
           </div>
         </div>
-        <div className="w-full h-0 border-b-[0.5px] border-b-gray-200 my-2"></div>
-        <div className={cn(rowClass)}>
-          <div className={labelClass}>
-            {t('datasetSettings.form.indexMethod')}
-          </div>
-          <div className='grow'>
-            <IndexMethodRadio
-              disable={!localeCurrentDataset?.embedding_available}
-              value={indexMethod}
-              onChange={v => setIndexMethod(v!)}
-              itemClassName='sm:!w-[280px]'
-            />
+        {currentDataset && currentDataset.indexing_technique && (
+          <div className={cn(rowClass)}>
+            <div className={labelClass}>
+              <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.indexMethod')}</div>
+            </div>
+            <div className='grow'>
+              <IndexMethodRadio
+                disable={!localeCurrentDataset?.embedding_available}
+                value={indexMethod}
+                onChange={v => setIndexMethod(v!)}
+                itemClassName='sm:!w-[280px]'
+              />
+            </div>
           </div>
-        </div>
+        )}
         {indexMethod === 'high_quality' && (
           <div className={cn(rowClass)}>
             <div className={labelClass}>
-              {t('datasetSettings.form.embeddingModel')}
+              <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.embeddingModel')}</div>
             </div>
             <div className='w-full'>
               <div className='w-full h-9 rounded-lg bg-gray-100 opacity-60'>
@@ -258,32 +281,75 @@ const SettingsModal: FC<SettingsModalProps> = ({
         )}
 
         {/* Retrieval Method Config */}
-        <div className={rowClass}>
-          <div className={cn(labelClass, 'w-auto min-w-[168px]')}>
-            <div>
-              <div>{t('datasetSettings.form.retrievalSetting.title')}</div>
-              <div className='leading-[18px] text-xs font-normal text-gray-500'>
-                <a target='_blank' rel='noopener noreferrer' href='https://docs.dify.ai/guides/knowledge-base/create-knowledge-and-upload-documents#id-4-retrieval-settings' className='text-[#155eef]'>{t('datasetSettings.form.retrievalSetting.learnMore')}</a>
-                {t('datasetSettings.form.retrievalSetting.description')}
+        {currentDataset?.provider === 'external'
+          ? <>
+            <div className={rowClass}><Divider/></div>
+            <div className={rowClass}>
+              <div className={labelClass}>
+                <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.retrievalSetting.title')}</div>
               </div>
+              <RetrievalSettings
+                topK={topK}
+                scoreThreshold={scoreThreshold}
+                scoreThresholdEnabled={scoreThresholdEnabled}
+                onChange={handleSettingsChange}
+                isInRetrievalSetting={true}
+              />
             </div>
-          </div>
-          <div>
-            {indexMethod === 'high_quality'
-              ? (
-                <RetrievalMethodConfig
-                  value={retrievalConfig}
-                  onChange={setRetrievalConfig}
-                />
-              )
-              : (
-                <EconomicalRetrievalMethodConfig
-                  value={retrievalConfig}
-                  onChange={setRetrievalConfig}
-                />
-              )}
-          </div>
-        </div>
+            <div className={rowClass}><Divider/></div>
+            <div className={rowClass}>
+              <div className={labelClass}>
+                <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.externalKnowledgeAPI')}</div>
+              </div>
+              <div className='w-full max-w-[480px]'>
+                <div className='flex h-full px-3 py-2 items-center gap-1 rounded-lg bg-components-input-bg-normal'>
+                  <ApiConnectionMod className='w-4 h-4 text-text-secondary' />
+                  <div className='overflow-hidden text-text-secondary text-ellipsis system-sm-medium'>
+                    {currentDataset?.external_knowledge_info.external_knowledge_api_name}
+                  </div>
+                  <div className='text-text-tertiary system-xs-regular'>·</div>
+                  <div className='text-text-tertiary system-xs-regular'>{currentDataset?.external_knowledge_info.external_knowledge_api_endpoint}</div>
+                </div>
+              </div>
+            </div>
+            <div className={rowClass}>
+              <div className={labelClass}>
+                <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.externalKnowledgeID')}</div>
+              </div>
+              <div className='w-full max-w-[480px]'>
+                <div className='flex h-full px-3 py-2 items-center gap-1 rounded-lg bg-components-input-bg-normal'>
+                  <div className='text-text-tertiary system-xs-regular'>{currentDataset?.external_knowledge_info.external_knowledge_id}</div>
+                </div>
+              </div>
+            </div>
+            <div className={rowClass}><Divider/></div>
+          </>
+          : <div className={rowClass}>
+            <div className={cn(labelClass, 'w-auto min-w-[168px]')}>
+              <div>
+                <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.retrievalSetting.title')}</div>
+                <div className='leading-[18px] text-xs font-normal text-gray-500'>
+                  <a target='_blank' rel='noopener noreferrer' href='https://docs.dify.ai/guides/knowledge-base/create-knowledge-and-upload-documents#id-4-retrieval-settings' className='text-[#155eef]'>{t('datasetSettings.form.retrievalSetting.learnMore')}</a>
+                  {t('datasetSettings.form.retrievalSetting.description')}
+                </div>
+              </div>
+            </div>
+            <div>
+              {indexMethod === 'high_quality'
+                ? (
+                  <RetrievalMethodConfig
+                    value={retrievalConfig}
+                    onChange={setRetrievalConfig}
+                  />
+                )
+                : (
+                  <EconomicalRetrievalMethodConfig
+                    value={retrievalConfig}
+                    onChange={setRetrievalConfig}
+                  />
+                )}
+            </div>
+          </div>}
       </div>
       {isRetrievalChanged && !isHideChangedTip && (
         <div className='absolute z-10 left-[30px] right-[30px] bottom-[76px] flex h-10 items-center px-3 rounded-lg border border-[#FEF0C7] bg-[#FFFAEB] shadow-lg justify-between'>

web/app/components/base/corner-label/index.tsx

@@ -0,0 +1,21 @@
+import { Corner } from '../icons/src/vender/solid/shapes'
+import cn from '@/utils/classnames'
+
+type CornerLabelProps = {
+  label: string
+  className?: string
+  labelClassName?: string
+}
+
+const CornerLabel: React.FC<CornerLabelProps> = ({ label, className, labelClassName }) => {
+  return (
+    <div className={cn('group/corner-label inline-flex items-start', className)}>
+      <Corner className='w-[13px] h-5 text-background-section group-hover/corner-label:text-background-section-burn' />
+      <div className={cn('flex py-1 pr-2 items-center gap-0.5 bg-background-section group-hover/corner-label:bg-background-section-burn', labelClassName)}>
+        <div className='text-text-tertiary system-2xs-medium-uppercase'>{label}</div>
+      </div>
+    </div>
+  )
+}
+
+export default CornerLabel

web/app/components/base/icons/assets/vender/solid/development/api-connection-mod.svg

@@ -0,0 +1,5 @@
+<svg width="16" height="16" viewBox="0 0 16 16" fill="none" xmlns="http://www.w3.org/2000/svg">
+<g id="Icon L">
+<path id="Vector" fill-rule="evenodd" clip-rule="evenodd" d="M7.99996 3.33333C5.42263 3.33333 3.33329 5.42267 3.33329 8C3.33329 10.5773 5.42263 12.6667 7.99996 12.6667C9.72643 12.6667 11.2348 11.7295 12.0427 10.3329C12.227 10.0141 12.6349 9.90523 12.9536 10.0896C13.2723 10.274 13.3812 10.6818 13.1968 11.0005C12.1604 12.7921 10.2216 14 7.99996 14C4.91159 14 2.36821 11.6666 2.03658 8.66667H1.33329C0.965103 8.66667 0.666626 8.36819 0.666626 8C0.666626 7.63181 0.965103 7.33333 1.33329 7.33333H2.03658C2.36821 4.33337 4.91159 2 7.99996 2C10.2216 2 12.1604 3.20785 13.1968 4.99952C13.3812 5.31823 13.2723 5.72605 12.9536 5.91041C12.6349 6.09477 12.227 5.98585 12.0427 5.66714C11.2348 4.27054 9.72643 3.33333 7.99996 3.33333ZM7.99996 6C6.89539 6 5.99996 6.89543 5.99996 8C5.99996 9.10455 6.89539 10 7.99996 10C9.1045 10 9.99996 9.10454 9.99996 8C9.99996 6.89543 9.10451 6 7.99996 6ZM4.66663 8C4.66663 6.15905 6.15901 4.66667 7.99996 4.66667C9.61257 4.66667 10.9578 5.81184 11.2666 7.33333H14.6666C15.0348 7.33333 15.3333 7.63181 15.3333 8C15.3333 8.36819 15.0348 8.66667 14.6666 8.66667H11.2666C10.9578 10.1881 9.61257 11.3333 7.99996 11.3333C6.159 11.3333 4.66663 9.84092 4.66663 8Z" fill="#354052"/>
+</g>
+</svg>

web/app/components/base/icons/assets/vender/solid/shapes/corner.svg

@@ -0,0 +1,3 @@
+<svg width="13" height="20" viewBox="0 0 13 20" fill="none" xmlns="http://www.w3.org/2000/svg">
+<path id="Shape" d="M0 0H13V20C9.98017 20 7.26458 18.1615 6.14305 15.3576L0 0Z" fill="#F9FAFB"/>
+</svg>

web/app/components/base/icons/src/vender/solid/development/ApiConnectionMod.json

@@ -0,0 +1,38 @@
+{
+	"icon": {
+		"type": "element",
+		"isRootNode": true,
+		"name": "svg",
+		"attributes": {
+			"width": "16",
+			"height": "16",
+			"viewBox": "0 0 16 16",
+			"fill": "none",
+			"xmlns": "http://www.w3.org/2000/svg"
+		},
+		"children": [
+			{
+				"type": "element",
+				"name": "g",
+				"attributes": {
+					"id": "Icon L"
+				},
+				"children": [
+					{
+						"type": "element",
+						"name": "path",
+						"attributes": {
+							"id": "Vector",
+							"fill-rule": "evenodd",
+							"clip-rule": "evenodd",
+							"d": "M7.99996 3.33333C5.42263 3.33333 3.33329 5.42267 3.33329 8C3.33329 10.5773 5.42263 12.6667 7.99996 12.6667C9.72643 12.6667 11.2348 11.7295 12.0427 10.3329C12.227 10.0141 12.6349 9.90523 12.9536 10.0896C13.2723 10.274 13.3812 10.6818 13.1968 11.0005C12.1604 12.7921 10.2216 14 7.99996 14C4.91159 14 2.36821 11.6666 2.03658 8.66667H1.33329C0.965103 8.66667 0.666626 8.36819 0.666626 8C0.666626 7.63181 0.965103 7.33333 1.33329 7.33333H2.03658C2.36821 4.33337 4.91159 2 7.99996 2C10.2216 2 12.1604 3.20785 13.1968 4.99952C13.3812 5.31823 13.2723 5.72605 12.9536 5.91041C12.6349 6.09477 12.227 5.98585 12.0427 5.66714C11.2348 4.27054 9.72643 3.33333 7.99996 3.33333ZM7.99996 6C6.89539 6 5.99996 6.89543 5.99996 8C5.99996 9.10455 6.89539 10 7.99996 10C9.1045 10 9.99996 9.10454 9.99996 8C9.99996 6.89543 9.10451 6 7.99996 6ZM4.66663 8C4.66663 6.15905 6.15901 4.66667 7.99996 4.66667C9.61257 4.66667 10.9578 5.81184 11.2666 7.33333H14.6666C15.0348 7.33333 15.3333 7.63181 15.3333 8C15.3333 8.36819 15.0348 8.66667 14.6666 8.66667H11.2666C10.9578 10.1881 9.61257 11.3333 7.99996 11.3333C6.159 11.3333 4.66663 9.84092 4.66663 8Z",
+							"fill": "currentColor"
+						},
+						"children": []
+					}
+				]
+			}
+		]
+	},
+	"name": "ApiConnectionMod"
+}

web/app/components/base/icons/src/vender/solid/development/ApiConnectionMod.tsx

@@ -0,0 +1,16 @@
+// GENERATE BY script
+// DON NOT EDIT IT MANUALLY
+
+import * as React from 'react'
+import data from './ApiConnectionMod.json'
+import IconBase from '@/app/components/base/icons/IconBase'
+import type { IconBaseProps, IconData } from '@/app/components/base/icons/IconBase'
+
+const Icon = React.forwardRef<React.MutableRefObject<SVGElement>, Omit<IconBaseProps, 'data'>>((
+  props,
+  ref,
+) => <IconBase {...props} ref={ref} data={data as IconData} />)
+
+Icon.displayName = 'ApiConnectionMod'
+
+export default Icon

web/app/components/base/icons/src/vender/solid/development/index.ts

@@ -1,3 +1,4 @@
+export { default as ApiConnectionMod } from './ApiConnectionMod'
 export { default as ApiConnection } from './ApiConnection'
 export { default as BarChartSquare02 } from './BarChartSquare02'
 export { default as Container } from './Container'

web/app/components/base/icons/src/vender/solid/shapes/Corner.json

@@ -0,0 +1,27 @@
+{
+	"icon": {
+		"type": "element",
+		"isRootNode": true,
+		"name": "svg",
+		"attributes": {
+			"width": "13",
+			"height": "20",
+			"viewBox": "0 0 13 20",
+			"fill": "none",
+			"xmlns": "http://www.w3.org/2000/svg"
+		},
+		"children": [
+			{
+				"type": "element",
+				"name": "path",
+				"attributes": {
+					"id": "Shape",
+					"d": "M0 0H13V20C9.98017 20 7.26458 18.1615 6.14305 15.3576L0 0Z",
+					"fill": "currentColor"
+				},
+				"children": []
+			}
+		]
+	},
+	"name": "Corner"
+}

web/app/components/base/icons/src/vender/solid/shapes/Corner.tsx

@@ -0,0 +1,16 @@
+// GENERATE BY script
+// DON NOT EDIT IT MANUALLY
+
+import * as React from 'react'
+import data from './Corner.json'
+import IconBase from '@/app/components/base/icons/IconBase'
+import type { IconBaseProps, IconData } from '@/app/components/base/icons/IconBase'
+
+const Icon = React.forwardRef<React.MutableRefObject<SVGElement>, Omit<IconBaseProps, 'data'>>((
+  props,
+  ref,
+) => <IconBase {...props} ref={ref} data={data as IconData} />)
+
+Icon.displayName = 'Corner'
+
+export default Icon

web/app/components/base/icons/src/vender/solid/shapes/index.ts

@@ -1,2 +1,3 @@
+export { default as Corner } from './Corner'
 export { default as Star04 } from './Star04'
 export { default as Star06 } from './Star06'

web/app/components/base/modal/index.css

@@ -3,5 +3,5 @@
 }
 
 .modal-panel {
-  @apply w-full max-w-md transform rounded-2xl bg-white p-6 text-left align-middle shadow-xl transition-all;
+  @apply w-full max-w-[480px] transform rounded-2xl bg-white p-6 text-left align-middle shadow-xl transition-all;
 }

web/app/components/base/param-item/index.tsx

@@ -37,6 +37,7 @@ const ParamItem: FC<Props> = ({ className, id, name, noTooltip, tip, step = 0.1,
           <span className="mx-1 text-gray-900 text-[13px] leading-[18px] font-medium">{name}</span>
           {!noTooltip && (
             <Tooltip
+              triggerClassName='w-4 h-4 shrink-0'
               popupContent={<div className="w-[200px]">{tip}</div>}
             />
           )}

web/app/components/base/select/index.tsx

@@ -87,7 +87,7 @@ const Select: FC<ISelectProps> = ({
         <div className='group text-gray-800'>
           {allowSearch
             ? <Combobox.Input
-              className={`w-full rounded-lg border-0 ${bgClassName} py-1.5 pl-3 pr-10 shadow-sm sm:text-sm sm:leading-6 focus-visible:outline-none focus-visible:bg-gray-200 group-hover:bg-gray-200 cursor-not-allowed`}
+              className={`w-full rounded-lg border-0 ${bgClassName} py-1.5 pl-3 pr-10 shadow-sm sm:text-sm sm:leading-6 focus-visible:outline-none focus-visible:bg-gray-200 group-hover:bg-gray-200 ${disabled ? 'cursor-not-allowed' : 'cursor-pointer'}`}
               onChange={(event) => {
                 if (!disabled)
                   setQuery(event.target.value)

web/app/components/datasets/documents/detail/completed/SegmentCard.tsx

@@ -36,6 +36,12 @@ export type UsageScene = 'doc' | 'hitTesting'
 type ISegmentCardProps = {
   loading: boolean
   detail?: SegmentDetailModel & { document: { name: string } }
+  contentExternal?: string
+  refSource?: {
+    title: string
+    uri: string
+  }
+  isExternal?: boolean
   score?: number
   onClick?: () => void
   onChangeSwitch?: (segId: string, enabled: boolean) => Promise<void>
@@ -48,6 +54,9 @@ type ISegmentCardProps = {
 
 const SegmentCard: FC<ISegmentCardProps> = ({
   detail = {},
+  contentExternal,
+  isExternal,
+  refSource,
   score,
   onClick,
   onChangeSwitch,
@@ -88,6 +97,9 @@ const SegmentCard: FC<ISegmentCardProps> = ({
       )
     }
 
+    if (contentExternal)
+      return contentExternal
+
     return content
   }
 
@@ -199,16 +211,16 @@ const SegmentCard: FC<ISegmentCardProps> = ({
               </div>
               <div className={cn('w-full bg-gray-50 group-hover:bg-white')}>
                 <Divider />
-                <div className="relative flex items-center w-full">
+                <div className="relative flex items-center w-full pb-1">
                   <DocumentTitle
-                    name={detail?.document?.name || ''}
-                    extension={(detail?.document?.name || '').split('.').pop() || 'txt'}
+                    name={detail?.document?.name || refSource?.title || ''}
+                    extension={(detail?.document?.name || refSource?.title || '').split('.').pop() || 'txt'}
                     wrapperCls='w-full'
                     iconCls="!h-4 !w-4 !bg-contain"
                     textCls="text-xs text-gray-700 !font-normal overflow-hidden whitespace-nowrap text-ellipsis"
                   />
                   <div className={cn(s.chartLinkText, 'group-hover:inline-flex')}>
-                    {t('datasetHitTesting.viewChart')}
+                    {isExternal ? t('datasetHitTesting.viewDetail') : t('datasetHitTesting.viewChart')}
                     <ArrowUpRightIcon className="w-3 h-3 ml-1 stroke-current stroke-2" />
                   </div>
                 </div>

web/app/components/datasets/external-api/declarations.ts

@@ -0,0 +1,16 @@
+export type CreateExternalAPIReq = {
+  name: string
+  settings: {
+    endpoint: string
+    api_key: string
+  }
+}
+
+export type FormSchema = {
+  variable: string
+  type: 'text' | 'secret'
+  label: {
+    [key: string]: string
+  }
+  required: boolean
+}

web/app/components/datasets/external-api/external-api-modal/Form.tsx

@@ -0,0 +1,90 @@
+import React, { useState } from 'react'
+import type { FC } from 'react'
+import { useTranslation } from 'react-i18next'
+import { RiBookOpenLine } from '@remixicon/react'
+import type { CreateExternalAPIReq, FormSchema } from '../declarations'
+import Input from '@/app/components/base/input'
+import cn from '@/utils/classnames'
+
+type FormProps = {
+  className?: string
+  itemClassName?: string
+  fieldLabelClassName?: string
+  value: CreateExternalAPIReq
+  onChange: (val: CreateExternalAPIReq) => void
+  formSchemas: FormSchema[]
+  inputClassName?: string
+}
+
+const Form: FC<FormProps> = React.memo(({
+  className,
+  itemClassName,
+  fieldLabelClassName,
+  value,
+  onChange,
+  formSchemas,
+  inputClassName,
+}) => {
+  const { t, i18n } = useTranslation()
+  const [changeKey, setChangeKey] = useState('')
+
+  const handleFormChange = (key: string, val: string) => {
+    setChangeKey(key)
+    if (key === 'name') {
+      onChange({ ...value, [key]: val })
+    }
+    else {
+      onChange({
+        ...value,
+        settings: {
+          ...value.settings,
+          [key]: val,
+        },
+      })
+    }
+  }
+
+  const renderField = (formSchema: FormSchema) => {
+    const { variable, type, label, required } = formSchema
+    const fieldValue = variable === 'name' ? value[variable] : (value.settings[variable as keyof typeof value.settings] || '')
+
+    return (
+      <div key={variable} className={cn(itemClassName, 'flex flex-col items-start gap-1 self-stretch')}>
+        <div className="flex justify-between items-center w-full">
+          <label className={cn(fieldLabelClassName, 'text-text-secondary system-sm-semibold')} htmlFor={variable}>
+            {label[i18n.language] || label.en_US}
+            {required && <span className='ml-1 text-red-500'>*</span>}
+          </label>
+          {variable === 'endpoint' && (
+            <a
+              href={'https://docs.dify.ai/guides/knowledge-base/external-knowledge-api-documentation' || '/'}
+              target='_blank'
+              rel='noopener noreferrer'
+              className='text-text-accent body-xs-regular flex items-center'
+            >
+              <RiBookOpenLine className='w-3 h-3 text-text-accent mr-1' />
+              {t('dataset.externalAPIPanelDocumentation')}
+            </a>
+          )}
+        </div>
+        <Input
+          type={type === 'secret' ? 'password' : 'text'}
+          id={variable}
+          name={variable}
+          value={fieldValue}
+          onChange={val => handleFormChange(variable, val.target.value)}
+          required={required}
+          className={cn(inputClassName)}
+        />
+      </div>
+    )
+  }
+
+  return (
+    <form className={cn('flex flex-col justify-center items-start gap-4 self-stretch', className)}>
+      {formSchemas.map(formSchema => renderField(formSchema))}
+    </form>
+  )
+})
+
+export default Form

web/app/components/datasets/external-api/external-api-modal/index.tsx

@@ -0,0 +1,218 @@
+import type { FC } from 'react'
+import {
+  memo,
+  useEffect,
+  useState,
+} from 'react'
+import { useTranslation } from 'react-i18next'
+import {
+  RiBook2Line,
+  RiCloseLine,
+  RiInformation2Line,
+  RiLock2Fill,
+} from '@remixicon/react'
+import type { CreateExternalAPIReq, FormSchema } from '../declarations'
+import Form from './Form'
+import ActionButton from '@/app/components/base/action-button'
+import Confirm from '@/app/components/base/confirm'
+import {
+  PortalToFollowElem,
+  PortalToFollowElemContent,
+} from '@/app/components/base/portal-to-follow-elem'
+import { createExternalAPI } from '@/service/datasets'
+import { useToastContext } from '@/app/components/base/toast'
+import Button from '@/app/components/base/button'
+import Tooltip from '@/app/components/base/tooltip'
+
+type AddExternalAPIModalProps = {
+  data?: CreateExternalAPIReq
+  onSave: (formValue: CreateExternalAPIReq) => void
+  onCancel: () => void
+  onEdit?: (formValue: CreateExternalAPIReq) => Promise<void>
+  datasetBindings?: { id: string; name: string }[]
+  isEditMode: boolean
+}
+
+const formSchemas: FormSchema[] = [
+  {
+    variable: 'name',
+    type: 'text',
+    label: {
+      en_US: 'Name',
+    },
+    required: true,
+  },
+  {
+    variable: 'endpoint',
+    type: 'text',
+    label: {
+      en_US: 'API Endpoint',
+    },
+    required: true,
+  },
+  {
+    variable: 'api_key',
+    type: 'secret',
+    label: {
+      en_US: 'API Key',
+    },
+    required: true,
+  },
+]
+
+const AddExternalAPIModal: FC<AddExternalAPIModalProps> = ({ data, onSave, onCancel, datasetBindings, isEditMode, onEdit }) => {
+  const { t } = useTranslation()
+  const { notify } = useToastContext()
+  const [loading, setLoading] = useState(false)
+  const [showConfirm, setShowConfirm] = useState(false)
+  const [formData, setFormData] = useState<CreateExternalAPIReq>({ name: '', settings: { endpoint: '', api_key: '' } })
+
+  useEffect(() => {
+    if (isEditMode && data)
+      setFormData(data)
+  }, [isEditMode, data])
+
+  const hasEmptyInputs = Object.values(formData).some(value =>
+    typeof value === 'string' ? value.trim() === '' : Object.values(value).some(v => v.trim() === ''),
+  )
+  const handleDataChange = (val: CreateExternalAPIReq) => {
+    setFormData(val)
+  }
+
+  const handleSave = async () => {
+    if (formData && formData.settings.api_key && formData.settings.api_key?.length < 5) {
+      notify({ type: 'error', message: t('common.apiBasedExtension.modal.apiKey.lengthError') })
+      setLoading(false)
+      return
+    }
+    try {
+      setLoading(true)
+      if (isEditMode && onEdit) {
+        await onEdit(
+          {
+            ...formData,
+            settings: { ...formData.settings, api_key: formData.settings.api_key ? '[__HIDDEN__]' : formData.settings.api_key },
+          },
+        )
+        notify({ type: 'success', message: 'External API updated successfully' })
+      }
+      else {
+        const res = await createExternalAPI({ body: formData })
+        if (res && res.id) {
+          notify({ type: 'success', message: 'External API saved successfully' })
+          onSave(res)
+        }
+      }
+      onCancel()
+    }
+    catch (error) {
+      console.error('Error saving/updating external API:', error)
+      notify({ type: 'error', message: 'Failed to save/update External API' })
+    }
+    finally {
+      setLoading(false)
+    }
+  }
+
+  return (
+    <PortalToFollowElem open>
+      <PortalToFollowElemContent className='w-full h-full z-[60]'>
+        <div className='fixed inset-0 flex items-center justify-center bg-black/[.25]'>
+          <div className='flex relative w-[480px] flex-col items-start bg-components-panel-bg rounded-2xl border-[0.5px] border-components-panel-border shadows-shadow-xl'>
+            <div className='flex flex-col pt-6 pl-6 pb-3 pr-14 items-start gap-2 self-stretch'>
+              <div className='self-stretch text-text-primary title-2xl-semi-bold flex-grow'>
+                {
+                  isEditMode ? t('dataset.editExternalAPIFormTitle') : t('dataset.createExternalAPI')
+                }
+              </div>
+              {isEditMode && (datasetBindings?.length ?? 0) > 0 && (
+                <div className='text-text-tertiary system-xs-regular flex items-center'>
+                  {t('dataset.editExternalAPIFormWarning.front')}
+                  <span className='text-text-accent cursor-pointer flex items-center'>
+                    &nbsp;{datasetBindings?.length} {t('dataset.editExternalAPIFormWarning.end')}&nbsp;
+                    <Tooltip
+                      popupClassName='flex items-center self-stretch w-[320px]'
+                      popupContent={
+                        <div className='p-1'>
+                          <div className='flex pt-1 pb-0.5 pl-2 pr-3 items-start self-stretch'>
+                            <div className='text-text-tertiary system-xs-medium-uppercase'>{`${datasetBindings?.length} ${t('dataset.editExternalAPITooltipTitle')}`}</div>
+                          </div>
+                          {datasetBindings?.map(binding => (
+                            <div key={binding.id} className='flex px-2 py-1 items-center gap-1 self-stretch'>
+                              <RiBook2Line className='w-4 h-4 text-text-secondary' />
+                              <div className='text-text-secondary system-sm-medium'>{binding.name}</div>
+                            </div>
+                          ))}
+                        </div>
+                      }
+                      asChild={false}
+                      position='bottom'
+                    >
+                      <RiInformation2Line className='w-3.5 h-3.5' />
+                    </Tooltip>
+                  </span>
+                </div>
+              )}
+            </div>
+            <ActionButton className='absolute top-5 right-5' onClick={onCancel}>
+              <RiCloseLine className='w-[18px] h-[18px] text-text-tertiary flex-shrink-0' />
+            </ActionButton>
+            <Form
+              value={formData}
+              onChange={handleDataChange}
+              formSchemas={formSchemas}
+              className='flex px-6 py-3 flex-col justify-center items-start gap-4 self-stretch'
+            />
+            <div className='flex p-6 pt-5 justify-end items-center gap-2 self-stretch'>
+              <Button type='button' variant='secondary' onClick={onCancel}>
+                {t('dataset.externalAPIForm.cancel')}
+              </Button>
+              <Button
+                type='submit'
+                variant='primary'
+                onClick={() => {
+                  if (isEditMode && (datasetBindings?.length ?? 0) > 0)
+                    setShowConfirm(true)
+                  else if (isEditMode && onEdit)
+                    onEdit(formData)
+
+                  else
+                    handleSave()
+                }}
+                disabled={hasEmptyInputs || loading}
+              >
+                {t('dataset.externalAPIForm.save')}
+              </Button>
+            </div>
+            <div className='flex px-2 py-3 justify-center items-center gap-1 self-stretch rounded-b-2xl
+              border-t-[0.5px] border-divider-subtle bg-background-soft text-text-tertiary system-xs-regular'
+            >
+              <RiLock2Fill className='w-3 h-3 text-text-quaternary' />
+              {t('dataset.externalAPIForm.encrypted.front')}
+              <a
+                className='text-text-accent'
+                target='_blank' rel='noopener noreferrer'
+                href='https://pycryptodome.readthedocs.io/en/latest/src/cipher/oaep.html'
+              >
+                PKCS1_OAEP
+              </a>
+              {t('dataset.externalAPIForm.encrypted.end')}
+            </div>
+          </div>
+          {showConfirm && (datasetBindings?.length ?? 0) > 0 && (
+            <Confirm
+              isShow={showConfirm}
+              type='warning'
+              title='Warning'
+              content={`${t('dataset.editExternalAPIConfirmWarningContent.front')} ${datasetBindings?.length} ${t('dataset.editExternalAPIConfirmWarningContent.end')}`}
+              onCancel={() => setShowConfirm(false)}
+              onConfirm={handleSave}
+            />
+          )}
+        </div>
+      </PortalToFollowElemContent>
+    </PortalToFollowElem>
+  )
+}
+
+export default memo(AddExternalAPIModal)

web/app/components/datasets/external-api/external-api-panel/index.tsx

@@ -0,0 +1,90 @@
+import React from 'react'
+import {
+  RiAddLine,
+  RiBookOpenLine,
+  RiCloseLine,
+} from '@remixicon/react'
+import { useTranslation } from 'react-i18next'
+import ExternalKnowledgeAPICard from '../external-knowledge-api-card'
+import cn from '@/utils/classnames'
+import { useExternalKnowledgeApi } from '@/context/external-knowledge-api-context'
+import ActionButton from '@/app/components/base/action-button'
+import Button from '@/app/components/base/button'
+import Loading from '@/app/components/base/loading'
+import { useModalContext } from '@/context/modal-context'
+
+type ExternalAPIPanelProps = {
+  onClose: () => void
+}
+
+const ExternalAPIPanel: React.FC<ExternalAPIPanelProps> = ({ onClose }) => {
+  const { t } = useTranslation()
+  const { setShowExternalKnowledgeAPIModal } = useModalContext()
+  const { externalKnowledgeApiList, mutateExternalKnowledgeApis, isLoading } = useExternalKnowledgeApi()
+
+  const handleOpenExternalAPIModal = () => {
+    setShowExternalKnowledgeAPIModal({
+      payload: { name: '', settings: { endpoint: '', api_key: '' } },
+      datasetBindings: [],
+      onSaveCallback: () => {
+        mutateExternalKnowledgeApis()
+      },
+      onCancelCallback: () => {
+        mutateExternalKnowledgeApis()
+      },
+      isEditMode: false,
+    })
+  }
+
+  return (
+    <div
+      tabIndex={-1}
+      className={cn('absolute top-14 right-0 bottom-2 flex z-10 outline-none')}
+    >
+      <div
+        className={cn(
+          'relative flex flex-col w-[420px] bg-components-panel-bg-alt rounded-l-2xl h-full border border-components-panel-border',
+        )}
+      >
+        <div className='flex items-start self-stretch p-4 pb-0'>
+          <div className='flex flex-col items-start gap-1 flex-grow'>
+            <div className='self-stretch text-text-primary system-xl-semibold'>{t('dataset.externalAPIPanelTitle')}</div>
+            <div className='self-stretch text-text-tertiary body-xs-regular'>{t('dataset.externalAPIPanelDescription')}</div>
+            <a className='flex justify-center items-center gap-1 self-stretch cursor-pointer' href='https://docs.dify.ai/guides/knowledge-base/external-knowledge-api-documentation' target='_blank'>
+              <RiBookOpenLine className='w-3 h-3 text-text-accent' />
+              <div className='flex-grow text-text-accent body-xs-regular'>{t('dataset.externalAPIPanelDocumentation')}</div>
+            </a>
+          </div>
+          <div className='flex items-center'>
+            <ActionButton onClick={() => onClose()}>
+              <RiCloseLine className='w-4 h-4 text-text-tertiary' />
+            </ActionButton>
+          </div>
+        </div>
+        <div className='flex px-4 py-3 flex-col justify-center items-start gap-2 self-stretch'>
+          <Button
+            variant={'primary'}
+            className='flex justify-center items-center px-3 py-2 gap-0.5'
+            onClick={handleOpenExternalAPIModal}
+          >
+            <RiAddLine className='w-4 h-4 text-components-button-primary-text' />
+            <div className='text-components-button-primary-text system-sm-medium'>{t('dataset.createExternalAPI')}</div>
+          </Button>
+        </div>
+        <div className='flex py-0 px-4 flex-col items-start gap-1 flex-grow self-stretch'>
+          {isLoading
+            ? (
+              <Loading />
+            )
+            : (
+              externalKnowledgeApiList.map(api => (
+                <ExternalKnowledgeAPICard key={api.id} api={api} />
+              ))
+            )}
+        </div>
+      </div>
+    </div>
+  )
+}
+
+export default ExternalAPIPanel

web/app/components/datasets/external-api/external-knowledge-api-card/index.tsx

@@ -0,0 +1,151 @@
+import React, { useState } from 'react'
+import { useTranslation } from 'react-i18next'
+import {
+  RiDeleteBinLine,
+  RiEditLine,
+} from '@remixicon/react'
+import type { CreateExternalAPIReq } from '../declarations'
+import type { ExternalAPIItem } from '@/models/datasets'
+import { checkUsageExternalAPI, deleteExternalAPI, fetchExternalAPI, updateExternalAPI } from '@/service/datasets'
+import { ApiConnectionMod } from '@/app/components/base/icons/src/vender/solid/development'
+import { useExternalKnowledgeApi } from '@/context/external-knowledge-api-context'
+import { useModalContext } from '@/context/modal-context'
+import ActionButton from '@/app/components/base/action-button'
+import Confirm from '@/app/components/base/confirm'
+
+type ExternalKnowledgeAPICardProps = {
+  api: ExternalAPIItem
+}
+
+const ExternalKnowledgeAPICard: React.FC<ExternalKnowledgeAPICardProps> = ({ api }) => {
+  const { setShowExternalKnowledgeAPIModal } = useModalContext()
+  const [showConfirm, setShowConfirm] = useState(false)
+  const [isHovered, setIsHovered] = useState(false)
+  const [usageCount, setUsageCount] = useState(0)
+  const { mutateExternalKnowledgeApis } = useExternalKnowledgeApi()
+
+  const { t } = useTranslation()
+
+  const handleEditClick = async () => {
+    try {
+      const response = await fetchExternalAPI({ apiTemplateId: api.id })
+      const formValue: CreateExternalAPIReq = {
+        name: response.name,
+        settings: {
+          endpoint: response.settings.endpoint,
+          api_key: response.settings.api_key,
+        },
+      }
+
+      setShowExternalKnowledgeAPIModal({
+        payload: formValue,
+        onSaveCallback: () => {
+          mutateExternalKnowledgeApis()
+        },
+        onCancelCallback: () => {
+          mutateExternalKnowledgeApis()
+        },
+        isEditMode: true,
+        datasetBindings: response.dataset_bindings,
+        onEditCallback: async (updatedData: CreateExternalAPIReq) => {
+          try {
+            await updateExternalAPI({
+              apiTemplateId: api.id,
+              body: {
+                ...response,
+                name: updatedData.name,
+                settings: {
+                  ...response.settings,
+                  endpoint: updatedData.settings.endpoint,
+                  api_key: updatedData.settings.api_key,
+                },
+              },
+            })
+            mutateExternalKnowledgeApis()
+          }
+          catch (error) {
+            console.error('Error updating external knowledge API:', error)
+          }
+        },
+      })
+    }
+    catch (error) {
+      console.error('Error fetching external knowledge API data:', error)
+    }
+  }
+
+  const handleDeleteClick = async () => {
+    try {
+      const usage = await checkUsageExternalAPI({ apiTemplateId: api.id })
+      if (usage.is_using)
+        setUsageCount(usage.count)
+
+      setShowConfirm(true)
+    }
+    catch (error) {
+      console.error('Error checking external API usage:', error)
+    }
+  }
+
+  const handleConfirmDelete = async () => {
+    try {
+      const response = await deleteExternalAPI({ apiTemplateId: api.id })
+      if (response && response.result === 'success') {
+        setShowConfirm(false)
+        mutateExternalKnowledgeApis()
+      }
+      else {
+        console.error('Failed to delete external API')
+      }
+    }
+    catch (error) {
+      console.error('Error deleting external knowledge API:', error)
+    }
+  }
+
+  return (
+    <>
+      <div className={`flex p-2 pl-3 items-start self-stretch rounded-lg border-[0.5px] 
+        border-components-panel-border-subtle bg-components-panel-on-panel-item-bg 
+        shadows-shadow-xs ${isHovered ? 'bg-state-destructive-hover border-state-destructive-border' : ''}`}
+      >
+        <div className='flex py-1 flex-col justify-center items-start gap-1.5 flex-grow'>
+          <div className='flex items-center gap-1 self-stretch text-text-secondary'>
+            <ApiConnectionMod className='w-4 h-4' />
+            <div className='system-sm-medium'>{api.name}</div>
+          </div>
+          <div className='self-stretch text-text-tertiary system-xs-regular'>{api.settings.endpoint}</div>
+        </div>
+        <div className='flex items-start gap-1'>
+          <ActionButton onClick={handleEditClick}>
+            <RiEditLine className='w-4 h-4 text-text-tertiary hover:text-text-secondary' />
+          </ActionButton>
+          <ActionButton
+            className='hover:bg-state-destructive-hover'
+            onClick={handleDeleteClick}
+            onMouseEnter={() => setIsHovered(true)}
+            onMouseLeave={() => setIsHovered(false)}
+          >
+            <RiDeleteBinLine className='w-4 h-4 text-text-tertiary hover:text-text-destructive' />
+          </ActionButton>
+        </div>
+      </div>
+      {showConfirm && (
+        <Confirm
+          isShow={showConfirm}
+          title={`${t('dataset.deleteExternalAPIConfirmWarningContent.title.front')} ${api.name}${t('dataset.deleteExternalAPIConfirmWarningContent.title.end')}`}
+          content={
+            usageCount > 0
+              ? `${t('dataset.deleteExternalAPIConfirmWarningContent.content.front')} ${usageCount} ${t('dataset.deleteExternalAPIConfirmWarningContent.content.end')}`
+              : t('dataset.deleteExternalAPIConfirmWarningContent.noConnectionContent')
+          }
+          type='warning'
+          onConfirm={handleConfirmDelete}
+          onCancel={() => setShowConfirm(false)}
+        />
+      )}
+    </>
+  )
+}
+
+export default ExternalKnowledgeAPICard

web/app/components/datasets/external-knowledge-base/connector/index.tsx

@@ -0,0 +1,36 @@
+'use client'
+
+import React, { useState } from 'react'
+import { useRouter } from 'next/navigation'
+import { useToastContext } from '@/app/components/base/toast'
+import ExternalKnowledgeBaseCreate from '@/app/components/datasets/external-knowledge-base/create'
+import type { CreateKnowledgeBaseReq } from '@/app/components/datasets/external-knowledge-base/create/declarations'
+import { createExternalKnowledgeBase } from '@/service/datasets'
+
+const ExternalKnowledgeBaseConnector = () => {
+  const { notify } = useToastContext()
+  const [loading, setLoading] = useState(false)
+  const router = useRouter()
+
+  const handleConnect = async (formValue: CreateKnowledgeBaseReq) => {
+    try {
+      setLoading(true)
+      const result = await createExternalKnowledgeBase({ body: formValue })
+      if (result && result.id) {
+        notify({ type: 'success', message: 'External Knowledge Base Connected Successfully' })
+        router.back()
+      }
+      else { throw new Error('Failed to create external knowledge base') }
+    }
+    catch (error) {
+      console.error('Error creating external knowledge base:', error)
+      notify({ type: 'error', message: 'Failed to connect External Knowledge Base' })
+    }
+    setLoading(false)
+  }
+  return (
+    <ExternalKnowledgeBaseCreate onConnect={handleConnect} loading={loading} />
+  )
+}
+
+export default ExternalKnowledgeBaseConnector

web/app/components/datasets/external-knowledge-base/create/ExternalApiSelect.tsx

@@ -0,0 +1,110 @@
+import React, { useEffect, useState } from 'react'
+import {
+  RiAddLine,
+  RiArrowDownSLine,
+} from '@remixicon/react'
+import { useTranslation } from 'react-i18next'
+import { useRouter } from 'next/navigation'
+import { ApiConnectionMod } from '@/app/components/base/icons/src/vender/solid/development'
+import { useModalContext } from '@/context/modal-context'
+import { useExternalKnowledgeApi } from '@/context/external-knowledge-api-context'
+
+type ApiItem = {
+  value: string
+  name: string
+  url: string
+}
+
+type ExternalApiSelectProps = {
+  items: ApiItem[]
+  value?: string
+  onSelect: (item: ApiItem) => void
+}
+
+const ExternalApiSelect: React.FC<ExternalApiSelectProps> = ({ items, value, onSelect }) => {
+  const { t } = useTranslation()
+  const [isOpen, setIsOpen] = useState(false)
+  const [selectedItem, setSelectedItem] = useState<ApiItem | null>(
+    items.find(item => item.value === value) || null,
+  )
+  const { setShowExternalKnowledgeAPIModal } = useModalContext()
+  const { mutateExternalKnowledgeApis } = useExternalKnowledgeApi()
+  const router = useRouter()
+
+  useEffect(() => {
+    const newSelectedItem = items.find(item => item.value === value) || null
+    setSelectedItem(newSelectedItem)
+  }, [value, items])
+
+  const handleAddNewAPI = () => {
+    setShowExternalKnowledgeAPIModal({
+      payload: { name: '', settings: { endpoint: '', api_key: '' } },
+      onSaveCallback: async () => {
+        mutateExternalKnowledgeApis()
+        router.refresh()
+      },
+      onCancelCallback: () => {
+        mutateExternalKnowledgeApis()
+      },
+      isEditMode: false,
+    })
+  }
+
+  const handleSelect = (item: ApiItem) => {
+    setSelectedItem(item)
+    onSelect(item)
+    setIsOpen(false)
+  }
+
+  return (
+    <div className="relative w-full">
+      <div
+        className={`flex items-center justify-between cursor-pointer px-2 py-1 gap-0.5 self-stretch rounded-lg 
+        bg-components-input-bg-normal hover:bg-state-base-hover-alt ${isOpen && 'bg-state-base-hover-alt'}`}
+        onClick={() => setIsOpen(!isOpen)}
+      >
+        {selectedItem
+          ? (
+            <div className="flex p-1 items-center gap-2 self-stretch rounded-lg">
+              <ApiConnectionMod className='text-text-secondary w-4 h-4' />
+              <div className='flex items-center flex-grow'>
+                <span className='text-components-input-text-filled text-ellipsis system-sm-regular overflow-hidden'>{selectedItem.name}</span>
+              </div>
+            </div>
+          )
+          : (
+            <span className='text-components-input-text-placeholder system-sm-regular'>{t('dataset.selectExternalKnowledgeAPI.placeholder')}</span>
+          )}
+        <RiArrowDownSLine className={`w-4 h-4 text-text-quaternary transition-transform ${isOpen ? 'text-text-secondary' : ''}`} />
+      </div>
+      {isOpen && (
+        <div className="absolute z-10 w-full mt-1 bg-components-panel-bg-blur border rounded-xl shadow-lg">
+          {items.map(item => (
+            <div
+              key={item.value}
+              className="flex p-1 items-center cursor-pointer"
+              onClick={() => handleSelect(item)}
+            >
+              <div className="flex p-2 items-center gap-2 self-stretch rounded-lg hover:bg-state-base-hover w-full">
+                <ApiConnectionMod className='text-text-secondary w-4 h-4' />
+                <span className='text-text-secondary text-ellipsis system-sm-medium overflow-hidden flex-grow'>{item.name}</span>
+                <span className='text-text-tertiary overflow-hidden text-right text-ellipsis system-xs-regular'>{item.url}</span>
+              </div>
+            </div>
+          ))}
+          <div className='flex p-1 flex-col items-start self-stretch'>
+            <div
+              className='flex p-2 items-center gap-2 self-stretch rounded-lg cursor-pointer hover:bg-state-base-hover'
+              onClick={handleAddNewAPI}
+            >
+              <RiAddLine className='text-text-secondary w-4 h-4' />
+              <span className='flex-grow overflow-hidden text-text-secondary text-ellipsis system-sm-medium'>{t('dataset.createNewExternalAPI')}</span>
+            </div>
+          </div>
+        </div>
+      )}
+    </div>
+  )
+}
+
+export default ExternalApiSelect

web/app/components/datasets/external-knowledge-base/create/ExternalApiSelection.tsx

@@ -0,0 +1,96 @@
+'use client'
+
+import React, { useEffect, useState } from 'react'
+import { useTranslation } from 'react-i18next'
+import { RiAddLine } from '@remixicon/react'
+import { useRouter } from 'next/navigation'
+import ExternalApiSelect from './ExternalApiSelect'
+import Input from '@/app/components/base/input'
+import Button from '@/app/components/base/button'
+import { useModalContext } from '@/context/modal-context'
+import { useExternalKnowledgeApi } from '@/context/external-knowledge-api-context'
+
+type ExternalApiSelectionProps = {
+  external_knowledge_api_id: string
+  external_knowledge_id: string
+  onChange: (data: { external_knowledge_api_id?: string; external_knowledge_id?: string }) => void
+}
+
+const ExternalApiSelection: React.FC<ExternalApiSelectionProps> = ({ external_knowledge_api_id, external_knowledge_id, onChange }) => {
+  const { t } = useTranslation()
+  const router = useRouter()
+  const { externalKnowledgeApiList } = useExternalKnowledgeApi()
+  const [selectedApiId, setSelectedApiId] = useState(external_knowledge_api_id)
+  const { setShowExternalKnowledgeAPIModal } = useModalContext()
+  const { mutateExternalKnowledgeApis } = useExternalKnowledgeApi()
+
+  const apiItems = externalKnowledgeApiList.map(api => ({
+    value: api.id,
+    name: api.name,
+    url: api.settings.endpoint,
+  }))
+
+  useEffect(() => {
+    if (apiItems.length > 0) {
+      const newSelectedId = external_knowledge_api_id || apiItems[0].value
+      setSelectedApiId(newSelectedId)
+      if (newSelectedId !== external_knowledge_api_id)
+        onChange({ external_knowledge_api_id: newSelectedId, external_knowledge_id })
+    }
+  }, [apiItems, external_knowledge_api_id, external_knowledge_id, onChange])
+
+  const handleAddNewAPI = () => {
+    setShowExternalKnowledgeAPIModal({
+      payload: { name: '', settings: { endpoint: '', api_key: '' } },
+      onSaveCallback: async () => {
+        mutateExternalKnowledgeApis()
+        router.refresh()
+      },
+      onCancelCallback: () => {
+        mutateExternalKnowledgeApis()
+      },
+      isEditMode: false,
+    })
+  }
+
+  useEffect(() => {
+    if (!external_knowledge_api_id && apiItems.length > 0)
+      onChange({ external_knowledge_api_id: apiItems[0].value, external_knowledge_id })
+  }, [])
+
+  return (
+    <form className='flex flex-col gap-4 self-stretch'>
+      <div className='flex flex-col gap-1 self-stretch'>
+        <div className='flex flex-col self-stretch'>
+          <label className='text-text-secondary system-sm-semibold'>{t('dataset.externalAPIPanelTitle')}</label>
+        </div>
+        {apiItems.length > 0
+          ? <ExternalApiSelect
+            items={apiItems}
+            value={selectedApiId}
+            onSelect={(e) => {
+              setSelectedApiId(e.value)
+              onChange({ external_knowledge_api_id: e.value, external_knowledge_id })
+            }}
+          />
+          : <Button variant={'tertiary'} onClick={handleAddNewAPI} className='justify-start gap-0.5'>
+            <RiAddLine className='w-4 h-4 text-text-tertiary' />
+            <span className='text-text-tertiary system-sm-regular'>{t('dataset.noExternalKnowledge')}</span>
+          </Button>
+        }
+      </div>
+      <div className='flex flex-col gap-1 self-stretch'>
+        <div className='flex flex-col self-stretch'>
+          <label className='text-text-secondary system-sm-semibold'>{t('dataset.externalKnowledgeId')}</label>
+        </div>
+        <Input
+          value={external_knowledge_id}
+          onChange={e => onChange({ external_knowledge_id: e.target.value, external_knowledge_api_id })}
+          placeholder={t('dataset.externalKnowledgeIdPlaceholder') ?? ''}
+        />
+      </div>
+    </form>
+  )
+}
+
+export default ExternalApiSelection

web/app/components/datasets/external-knowledge-base/create/InfoPanel.tsx

@@ -0,0 +1,33 @@
+import { RiBookOpenLine } from '@remixicon/react'
+import { useTranslation } from 'react-i18next'
+
+const InfoPanel = () => {
+  const { t } = useTranslation()
+
+  return (
+    <div className='flex w-[360px] pt-[108px] pb-2 pr-8 flex-col items-start'>
+      <div className='flex min-w-[240px] w-full p-6 flex-col items-start gap-3 self-stretch rounded-xl bg-background-section'>
+        <div className='flex p-1 w-10 h-10 justify-center items-center gap-2 flex-grow self-stretch rounded-lg border-[0.5px] border-components-card-border bg-components-card-bg'>
+          <RiBookOpenLine className='w-5 h-5 text-text-accent' />
+        </div>
+        <p className='flex flex-col items-start gap-2 self-stretch'>
+          <span className='self-stretch text-text-secondary system-xl-semibold'>
+            {t('dataset.connectDatasetIntro.title')}
+          </span>
+          <span className='text-text-tertiary system-sm-regular'>
+            {t('dataset.connectDatasetIntro.content.front')}
+            <a className='text-text-accent system-sm-regular ml-1' href='https://docs.dify.ai/guides/knowledge-base/external-knowledge-api-documentation' target='_blank' rel="noopener noreferrer">
+              {t('dataset.connectDatasetIntro.content.link')}
+            </a>
+            {t('dataset.connectDatasetIntro.content.end')}
+          </span>
+          <a className='self-stretch text-text-accent system-sm-regular' href='https://docs.dify.ai/guides/knowledge-base/connect-external-knowledge' target='_blank' rel="noopener noreferrer">
+            {t('dataset.connectDatasetIntro.learnMore')}
+          </a>
+        </p>
+      </div>
+    </div>
+  )
+}
+
+export default InfoPanel

web/app/components/datasets/external-knowledge-base/create/KnowledgeBaseInfo.tsx

@@ -0,0 +1,65 @@
+import React from 'react'
+import { RiBookOpenLine } from '@remixicon/react'
+import { useTranslation } from 'react-i18next'
+import Input from '@/app/components/base/input'
+
+type KnowledgeBaseInfoProps = {
+  name: string
+  description?: string
+  onChange: (data: { name?: string; description?: string }) => void
+}
+
+const KnowledgeBaseInfo: React.FC<KnowledgeBaseInfoProps> = ({ name, description, onChange }) => {
+  const { t } = useTranslation()
+
+  const handleNameChange = (e: React.ChangeEvent<HTMLInputElement>) => {
+    onChange({ name: e.target.value })
+  }
+
+  const handleDescriptionChange = (e: React.ChangeEvent<HTMLTextAreaElement>) => {
+    onChange({ description: e.target.value })
+  }
+
+  return (
+    <form className='flex flex-col gap-4 self-stretch'>
+      <div className='flex flex-col gap-4 self-stretch'>
+        <div className='flex flex-col gap-1 self-stretch'>
+          <div className='flex flex-col justify-center self-stretch'>
+            <label className='text-text-secondary system-sm-semibold'>{t('dataset.externalKnowledgeName')}</label>
+          </div>
+          <Input
+            value={name}
+            onChange={handleNameChange}
+            placeholder={t('dataset.externalKnowledgeNamePlaceholder') ?? ''}
+          />
+        </div>
+        <div className='flex flex-col gap-1 self-stretch'>
+          <div className='flex flex-col justify-center self-stretch'>
+            <label className='text-text-secondary system-sm-semibold'>{t('dataset.externalKnowledgeDescription')}</label>
+          </div>
+          <div className='flex flex-col gap-1 self-stretch'>
+            <textarea
+              value={description}
+              onChange={ e => handleDescriptionChange(e)}
+              placeholder={t('dataset.externalKnowledgeDescriptionPlaceholder') ?? ''}
+              className={`flex h-20 py-2 p-3 self-stretch items-start rounded-lg bg-components-input-bg-normal ${description ? 'text-components-input-text-filled' : 'text-components-input-text-placeholder'} system-sm-regular`}
+            />
+            <a
+              className='flex py-0.5 gap-1 self-stretch'
+              href='https://docs.dify.ai/features/datasets#how-to-write-a-good-dataset-description'
+              target="_blank"
+              rel="noopener noreferrer"
+            >
+              <div className='flex p-0.5 items-center gap-2'>
+                <RiBookOpenLine className='w-3 h-3 text-text-tertiary' />
+              </div>
+              <div className='flex-grow text-text-tertiary body-xs-regular'>{t('dataset.learnHowToWriteGoodKnowledgeDescription')}</div>
+            </a>
+          </div>
+        </div>
+      </div>
+    </form>
+  )
+}
+
+export default KnowledgeBaseInfo

web/app/components/datasets/external-knowledge-base/create/RetrievalSettings.tsx

@@ -0,0 +1,67 @@
+import type { FC } from 'react'
+import React from 'react'
+import { useTranslation } from 'react-i18next'
+import TopKItem from '@/app/components/base/param-item/top-k-item'
+import ScoreThresholdItem from '@/app/components/base/param-item/score-threshold-item'
+import cn from '@/utils/classnames'
+
+type RetrievalSettingsProps = {
+  topK: number
+  scoreThreshold: number
+  scoreThresholdEnabled: boolean
+  isInHitTesting?: boolean
+  isInRetrievalSetting?: boolean
+  onChange: (data: { top_k?: number; score_threshold?: number; score_threshold_enabled?: boolean }) => void
+}
+
+const RetrievalSettings: FC<RetrievalSettingsProps> = ({
+  topK,
+  scoreThreshold,
+  scoreThresholdEnabled,
+  onChange,
+  isInHitTesting = false,
+  isInRetrievalSetting = false,
+}) => {
+  const { t } = useTranslation()
+
+  const handleScoreThresholdChange = (enabled: boolean) => {
+    onChange({ score_threshold_enabled: enabled })
+  }
+
+  return (
+    <div className={cn('flex flex-col gap-2 self-stretch', isInRetrievalSetting && 'w-full max-w-[480px]')}>
+      {!isInHitTesting && !isInRetrievalSetting && <div className='flex h-7 pt-1 flex-col gap-2 self-stretch'>
+        <label className='text-text-secondary system-sm-semibold'>{t('dataset.retrievalSettings')}</label>
+      </div>}
+      <div className={cn(
+        'flex gap-4 self-stretch',
+        {
+          'flex-col': isInHitTesting,
+          'flex-row': isInRetrievalSetting,
+          'flex-col sm:flex-row': !isInHitTesting && !isInRetrievalSetting,
+        },
+      )}>
+        <div className='flex flex-col gap-1 flex-grow'>
+          <TopKItem
+            className='grow'
+            value={topK}
+            onChange={(_key, v) => onChange({ top_k: v })}
+            enable={true}
+          />
+        </div>
+        <div className='flex flex-col gap-1 flex-grow'>
+          <ScoreThresholdItem
+            className='grow'
+            value={scoreThreshold}
+            onChange={(_key, v) => onChange({ score_threshold: v })}
+            enable={scoreThresholdEnabled}
+            hasSwitch={true}
+            onSwitchChange={(_key, v) => handleScoreThresholdChange(v)}
+          />
+        </div>
+      </div>
+    </div>
+  )
+}
+
+export default RetrievalSettings

web/app/components/datasets/external-knowledge-base/create/declarations.ts

@@ -0,0 +1,12 @@
+export type CreateKnowledgeBaseReq = {
+  name: string
+  description?: string
+  external_knowledge_api_id: string
+  provider: 'external'
+  external_knowledge_id: string
+  external_retrieval_model: {
+    top_k: number
+    score_threshold: number
+    score_threshold_enabled: boolean
+  }
+}

web/app/components/datasets/external-knowledge-base/create/index.tsx

@@ -0,0 +1,128 @@
+'use client'
+
+import { useCallback, useState } from 'react'
+import { useRouter } from 'next/navigation'
+import { RiArrowLeftLine, RiArrowRightLine } from '@remixicon/react'
+import { useTranslation } from 'react-i18next'
+import KnowledgeBaseInfo from './KnowledgeBaseInfo'
+import ExternalApiSelection from './ExternalApiSelection'
+import RetrievalSettings from './RetrievalSettings'
+import InfoPanel from './InfoPanel'
+import type { CreateKnowledgeBaseReq } from './declarations'
+import Divider from '@/app/components/base/divider'
+import Button from '@/app/components/base/button'
+
+type ExternalKnowledgeBaseCreateProps = {
+  onConnect: (formValue: CreateKnowledgeBaseReq) => void
+  loading: boolean
+}
+
+const ExternalKnowledgeBaseCreate: React.FC<ExternalKnowledgeBaseCreateProps> = ({ onConnect, loading }) => {
+  const { t } = useTranslation()
+  const router = useRouter()
+  const [formData, setFormData] = useState<CreateKnowledgeBaseReq>({
+    name: '',
+    description: '',
+    external_knowledge_api_id: '',
+    external_knowledge_id: '',
+    external_retrieval_model: {
+      top_k: 2,
+      score_threshold: 0.5,
+      score_threshold_enabled: false,
+    },
+    provider: 'external',
+
+  })
+
+  const navBackHandle = useCallback(() => {
+    router.replace('/datasets')
+  }, [router])
+
+  const handleFormChange = (newData: CreateKnowledgeBaseReq) => {
+    setFormData(newData)
+  }
+
+  const isFormValid = formData.name.trim() !== ''
+    && formData.external_knowledge_api_id !== ''
+    && formData.external_knowledge_id !== ''
+    && formData.external_retrieval_model.top_k !== undefined
+    && formData.external_retrieval_model.score_threshold !== undefined
+
+  return (
+    <div className='flex flex-col flex-grow self-stretch rounded-t-2xl border-t border-effects-highlight bg-components-panel-bg'>
+      <div className='flex justify-center flex-grow self-stretch'>
+        <div className='flex w-full max-w-[960px] px-14 py-0 flex-col items-center'>
+          <div className='flex w-full max-w-[640px] pt-6 pb-8 flex-col grow items-center gap-4'>
+            <div className='relative flex flex-col py-2 items-center gap-[2px] self-stretch'>
+              <div className='flex-grow text-text-primary system-xl-semibold self-stretch'>{t('dataset.connectDataset')}</div>
+              <p className='text-text-tertiary system-sm-regular'>
+                <span>{t('dataset.connectHelper.helper1')}</span>
+                <span className='text-text-secondary system-sm-medium'>{t('dataset.connectHelper.helper2')}</span>
+                <span>{t('dataset.connectHelper.helper3')}</span>
+                <a className='self-stretch text-text-accent system-sm-regular' href='https://docs.dify.ai/guides/knowledge-base/connect-external-knowledge' target='_blank' rel="noopener noreferrer">
+                  {t('dataset.connectHelper.helper4')}
+                </a>
+                <span>{t('dataset.connectHelper.helper5')} </span>
+              </p>
+              <Button
+                className='flex w-8 h-8 p-2 items-center justify-center absolute left-[-44px] top-1 rounded-full'
+                variant='tertiary'
+                onClick={navBackHandle}
+              >
+                <RiArrowLeftLine className='w-4 h-4 text-text-tertiary' />
+              </Button>
+            </div>
+            <KnowledgeBaseInfo
+              name={formData.name}
+              description={formData.description ?? ''}
+              onChange={data => handleFormChange({
+                ...formData,
+                ...data,
+              })}
+            />
+            <Divider />
+            <ExternalApiSelection
+              external_knowledge_api_id={formData.external_knowledge_api_id}
+              external_knowledge_id={formData.external_knowledge_id}
+              onChange={data => handleFormChange({
+                ...formData,
+                ...data,
+              })}
+            />
+            <RetrievalSettings
+              topK={formData.external_retrieval_model.top_k}
+              scoreThreshold={formData.external_retrieval_model.score_threshold}
+              scoreThresholdEnabled={formData.external_retrieval_model.score_threshold_enabled}
+              onChange={data => handleFormChange({
+                ...formData,
+                external_retrieval_model: {
+                  ...formData.external_retrieval_model,
+                  ...data,
+                },
+              })}
+            />
+            <div className='flex py-2 justify-end items-center gap-2 self-stretch'>
+              <Button variant='secondary' onClick={navBackHandle}>
+                <div className='text-components-button-secondary-text system-sm-medium'>{t('dataset.externalKnowledgeForm.cancel')}</div>
+              </Button>
+              <Button
+                variant='primary'
+                onClick={() => {
+                  onConnect(formData)
+                }}
+                disabled={!isFormValid}
+                loading={loading}
+              >
+                <div className='text-components-button-primary-text system-sm-medium'>{t('dataset.externalKnowledgeForm.connect')}</div>
+                <RiArrowRightLine className='w-4 h-4 text-components-button-primary-text' />
+              </Button>
+            </div>
+          </div>
+        </div>
+        <InfoPanel />
+      </div>
+    </div>
+  )
+}
+
+export default ExternalKnowledgeBaseCreate

web/app/components/datasets/hit-testing/hit-detail.tsx

@@ -26,12 +26,15 @@ const HitDetail: FC<IHitDetailProps> = ({ segInfo }) => {
       )
     }
 
-    return segInfo?.content
+    return <div className='mb-4 text-md text-gray-800 h-full'>{segInfo?.content}</div>
   }
 
   return (
-    <div className='overflow-x-auto'>
-      <div className="bg-gray-25 p-6">
+    segInfo?.id === 'external'
+      ? <div className='w-full overflow-x-auto px-2'>
+        <div className={s.segModalContent}>{renderContent()}</div>
+      </div>
+      : <div className='overflow-x-auto'>
         <div className="flex items-center">
           <SegmentIndexTag
             positionId={segInfo?.position || ''}
@@ -59,7 +62,6 @@ const HitDetail: FC<IHitDetailProps> = ({ segInfo }) => {
             })}
         </div>
       </div>
-    </div>
   )
 }
 

web/app/components/datasets/hit-testing/index.tsx

@@ -13,7 +13,7 @@ import s from './style.module.css'
 import HitDetail from './hit-detail'
 import ModifyRetrievalModal from './modify-retrieval-modal'
 import cn from '@/utils/classnames'
-import type { HitTestingResponse, HitTesting as HitTestingType } from '@/models/datasets'
+import type { ExternalKnowledgeBaseHitTestingResponse, ExternalKnowledgeBaseHitTesting as ExternalKnowledgeBaseHitTestingType, HitTestingResponse, HitTesting as HitTestingType } from '@/models/datasets'
 import Loading from '@/app/components/base/loading'
 import Modal from '@/app/components/base/modal'
 import Drawer from '@/app/components/base/drawer'
@@ -49,8 +49,10 @@ const HitTesting: FC<Props> = ({ datasetId }: Props) => {
   const isMobile = media === MediaType.mobile
 
   const [hitResult, setHitResult] = useState<HitTestingResponse | undefined>() // 初始化记录为空数组
+  const [externalHitResult, setExternalHitResult] = useState<ExternalKnowledgeBaseHitTestingResponse | undefined>()
   const [submitLoading, setSubmitLoading] = useState(false)
   const [currParagraph, setCurrParagraph] = useState<{ paraInfo?: HitTestingType; showModal: boolean }>({ showModal: false })
+  const [externalCurrParagraph, setExternalCurrParagraph] = useState<{ paraInfo?: ExternalKnowledgeBaseHitTestingType; showModal: boolean }>({ showModal: false })
   const [text, setText] = useState('')
 
   const [currPage, setCurrPage] = React.useState<number>(0)
@@ -66,12 +68,52 @@ const HitTesting: FC<Props> = ({ datasetId }: Props) => {
     setCurrParagraph({ paraInfo: detail, showModal: true })
   }
 
+  const onClickExternalCard = (detail: ExternalKnowledgeBaseHitTestingType) => {
+    setExternalCurrParagraph({ paraInfo: detail, showModal: true })
+  }
   const { dataset: currentDataset } = useContext(DatasetDetailContext)
+  const isExternal = currentDataset?.provider === 'external'
 
   const [retrievalConfig, setRetrievalConfig] = useState(currentDataset?.retrieval_model_dict as RetrievalConfig)
   const [isShowModifyRetrievalModal, setIsShowModifyRetrievalModal] = useState(false)
   const [isShowRightPanel, { setTrue: showRightPanel, setFalse: hideRightPanel, set: setShowRightPanel }] = useBoolean(!isMobile)
 
+  const renderHitResults = (results: any[], onClickCard: (record: any) => void) => (
+    <>
+      <div className='text-gray-600 font-semibold mb-4'>{t('datasetHitTesting.hit.title')}</div>
+      <div className='overflow-auto flex-1'>
+        <div className={s.cardWrapper}>
+          {results.map((record, idx) => (
+            <SegmentCard
+              key={idx}
+              loading={false}
+              refSource= {{
+                title: record.title,
+                uri: record.metadata ? record.metadata['x-amz-bedrock-kb-source-uri'] : '',
+              }}
+              isExternal={isExternal}
+              detail={record.segment}
+              contentExternal={record.content}
+              score={record.score}
+              scene='hitTesting'
+              className='h-[216px] mb-4'
+              onClick={() => onClickCard(record)}
+            />
+          ))}
+        </div>
+      </div>
+    </>
+  )
+
+  const renderEmptyState = () => (
+    <div className='h-full flex flex-col justify-center items-center'>
+      <div className={cn(docStyle.commonIcon, docStyle.targetIcon, '!bg-gray-200 !h-14 !w-14')} />
+      <div className='text-gray-300 text-[13px] mt-3'>
+        {t('datasetHitTesting.hit.emptyTip')}
+      </div>
+    </div>
+  )
+
   useEffect(() => {
     setShowRightPanel(!isMobile)
   }, [isMobile, setShowRightPanel])
@@ -86,12 +128,14 @@ const HitTesting: FC<Props> = ({ datasetId }: Props) => {
         <Textarea
           datasetId={datasetId}
           setHitResult={setHitResult}
+          setExternalHitResult={setExternalHitResult}
           onSubmit={showRightPanel}
           onUpdateList={recordsMutate}
           loading={submitLoading}
           setLoading={setSubmitLoading}
           setText={setText}
           text={text}
+          isExternal={isExternal}
           onClickRetrievalMethod={() => setIsShowModifyRetrievalModal(true)}
           retrievalConfig={retrievalConfig}
           isEconomy={currentDataset?.indexing_technique === 'economy'}
@@ -159,47 +203,42 @@ const HitTesting: FC<Props> = ({ datasetId }: Props) => {
                 className='h-[216px]'
               />
             </div>
-            : !hitResult?.records.length
-              ? (
-                <div className='h-full flex flex-col justify-center items-center'>
-                  <div className={cn(docStyle.commonIcon, docStyle.targetIcon, '!bg-gray-200 !h-14 !w-14')} />
-                  <div className='text-gray-300 text-[13px] mt-3'>
-                    {t('datasetHitTesting.hit.emptyTip')}
-                  </div>
-                </div>
-              )
-              : (
-                <>
-                  <div className='text-gray-600 font-semibold mb-4'>{t('datasetHitTesting.hit.title')}</div>
-                  <div className='overflow-auto flex-1'>
-                    <div className={s.cardWrapper}>
-                      {hitResult?.records.map((record, idx) => {
-                        return <SegmentCard
-                          key={idx}
-                          loading={false}
-                          detail={record.segment as any}
-                          score={record.score}
-                          scene='hitTesting'
-                          className='h-[216px] mb-4'
-                          onClick={() => onClickCard(record as any)}
-                        />
-                      })}
-                    </div>
-                  </div>
-                </>
-              )
+            : (
+              (() => {
+                if (!hitResult?.records.length && !externalHitResult?.records.length)
+                  return renderEmptyState()
+
+                if (hitResult?.records.length)
+                  return renderHitResults(hitResult.records, onClickCard)
+
+                return renderHitResults(externalHitResult?.records || [], onClickExternalCard)
+              })()
+            )
           }
         </div>
       </FloatRightContainer>
       <Modal
-        className='w-[520px] p-0'
+        className={isExternal ? 'py-10 px-8' : 'w-full'}
         closable
-        onClose={() => setCurrParagraph({ showModal: false })}
-        isShow={currParagraph.showModal}
+        onClose={() => {
+          setCurrParagraph({ showModal: false })
+          setExternalCurrParagraph({ showModal: false })
+        }}
+        isShow={currParagraph.showModal || externalCurrParagraph.showModal}
       >
-        {currParagraph.showModal && <HitDetail
-          segInfo={currParagraph.paraInfo?.segment}
-        />}
+        {currParagraph.showModal && (
+          <HitDetail
+            segInfo={currParagraph.paraInfo?.segment}
+          />
+        )}
+        {externalCurrParagraph.showModal && (
+          <HitDetail
+            segInfo={{
+              id: 'external',
+              content: externalCurrParagraph.paraInfo?.content,
+            }}
+          />
+        )}
       </Modal>
       <Drawer isOpen={isShowModifyRetrievalModal} onClose={() => setIsShowModifyRetrievalModal(false)} footer={null} mask={isMobile} panelClassname='mt-16 mx-2 sm:mr-2 mb-3 !p-0 !max-w-[640px] rounded-xl'>
         <ModifyRetrievalModal

web/app/components/datasets/hit-testing/modify-external-retrieval-modal.tsx

@@ -0,0 +1,71 @@
+import { useState } from 'react'
+import {
+  RiCloseLine,
+} from '@remixicon/react'
+import { useTranslation } from 'react-i18next'
+import RetrievalSettings from '../external-knowledge-base/create/RetrievalSettings'
+import Button from '@/app/components/base/button'
+import ActionButton from '@/app/components/base/action-button'
+
+type ModifyExternalRetrievalModalProps = {
+  onClose: () => void
+  onSave: (data: { top_k: number; score_threshold: number; score_threshold_enabled: boolean }) => void
+  initialTopK: number
+  initialScoreThreshold: number
+  initialScoreThresholdEnabled: boolean
+}
+
+const ModifyExternalRetrievalModal: React.FC<ModifyExternalRetrievalModalProps> = ({
+  onClose,
+  onSave,
+  initialTopK,
+  initialScoreThreshold,
+  initialScoreThresholdEnabled,
+}) => {
+  const { t } = useTranslation()
+  const [topK, setTopK] = useState(initialTopK)
+  const [scoreThreshold, setScoreThreshold] = useState(initialScoreThreshold)
+  const [scoreThresholdEnabled, setScoreThresholdEnabled] = useState(initialScoreThresholdEnabled)
+
+  const handleSettingsChange = (data: { top_k?: number; score_threshold?: number; score_threshold_enabled?: boolean }) => {
+    if (data.top_k !== undefined)
+      setTopK(data.top_k)
+    if (data.score_threshold !== undefined)
+      setScoreThreshold(data.score_threshold)
+    if (data.score_threshold_enabled !== undefined)
+      setScoreThresholdEnabled(data.score_threshold_enabled)
+  }
+
+  const handleSave = () => {
+    onSave({ top_k: topK, score_threshold: scoreThreshold, score_threshold_enabled: scoreThresholdEnabled })
+    onClose()
+  }
+
+  return (
+    <div className='absolute z-10 top-[36px] right-[14px] flex w-[320px] flex-col items-start rounded-2xl border-[0.5px]
+      border-components-panel-border bg-components-panel-bg shadows-shadow-2xl'
+    >
+      <div className='flex p-4 pb-2 items-center justify-between self-stretch'>
+        <div className='text-text-primary system-xl-semibold flex-grow'>{t('datasetHitTesting.settingTitle')}</div>
+        <ActionButton className='ml-auto' onClick={onClose}>
+          <RiCloseLine className='w-4 h-4 flex-shrink-0' />
+        </ActionButton>
+      </div>
+      <div className='flex p-4 pt-2 flex-col justify-center items-start gap-4 self-stretch'>
+        <RetrievalSettings
+          topK={topK}
+          scoreThreshold={scoreThreshold}
+          scoreThresholdEnabled={scoreThresholdEnabled}
+          onChange={handleSettingsChange}
+          isInHitTesting={true}
+        />
+      </div>
+      <div className='flex p-4 pt-2 justify-end items-end gap-1 w-full'>
+        <Button className='flex-shrink-0 min-w-[72px]' onClick={onClose}>{t('common.operation.cancel')}</Button>
+        <Button variant='primary' className='flex-shrink-0 min-w-[72px]' onClick={handleSave}>{t('common.operation.save')}</Button>
+      </div>
+    </div>
+  )
+}
+
+export default ModifyExternalRetrievalModal

web/app/components/datasets/hit-testing/textarea.tsx

@@ -1,12 +1,17 @@
+import React, { useState } from 'react'
 import { useTranslation } from 'react-i18next'
+import {
+  RiEqualizer2Line,
+} from '@remixicon/react'
 import Button from '../../base/button'
 import Tag from '../../base/tag'
 import { getIcon } from '../common/retrieval-method-info'
 import s from './style.module.css'
+import ModifyExternalRetrievalModal from './modify-external-retrieval-modal'
 import Tooltip from '@/app/components/base/tooltip'
 import cn from '@/utils/classnames'
-import type { HitTestingResponse } from '@/models/datasets'
-import { hitTesting } from '@/service/datasets'
+import type { ExternalKnowledgeBaseHitTestingResponse, HitTestingResponse } from '@/models/datasets'
+import { externalKnowledgeBaseHitTesting, hitTesting } from '@/service/datasets'
 import { asyncRunSafe } from '@/utils'
 import { RETRIEVE_METHOD, type RetrievalConfig } from '@/types/app'
 
@@ -14,10 +19,12 @@ type TextAreaWithButtonIProps = {
   datasetId: string
   onUpdateList: () => void
   setHitResult: (res: HitTestingResponse) => void
+  setExternalHitResult: (res: ExternalKnowledgeBaseHitTestingResponse) => void
   loading: boolean
   setLoading: (v: boolean) => void
   text: string
   setText: (v: string) => void
+  isExternal?: boolean
   onClickRetrievalMethod: () => void
   retrievalConfig: RetrievalConfig
   isEconomy: boolean
@@ -28,16 +35,29 @@ const TextAreaWithButton = ({
   datasetId,
   onUpdateList,
   setHitResult,
+  setExternalHitResult,
   setLoading,
   loading,
   text,
   setText,
+  isExternal = false,
   onClickRetrievalMethod,
   retrievalConfig,
   isEconomy,
   onSubmit: _onSubmit,
 }: TextAreaWithButtonIProps) => {
   const { t } = useTranslation()
+  const [isSettingsOpen, setIsSettingsOpen] = useState(false)
+  const [externalRetrievalSettings, setExternalRetrievalSettings] = useState({
+    top_k: 2,
+    score_threshold: 0.5,
+    score_threshold_enabled: false,
+  })
+
+  const handleSaveExternalRetrievalSettings = (data: { top_k: number; score_threshold: number; score_threshold_enabled: boolean }) => {
+    setExternalRetrievalSettings(data)
+    setIsSettingsOpen(false)
+  }
 
   function handleTextChange(event: any) {
     setText(event.target.value)
@@ -63,28 +83,70 @@ const TextAreaWithButton = ({
     _onSubmit && _onSubmit()
   }
 
+  const externalRetrievalTestingOnSubmit = async () => {
+    const [e, res] = await asyncRunSafe<ExternalKnowledgeBaseHitTestingResponse>(
+      externalKnowledgeBaseHitTesting({
+        datasetId,
+        query: text,
+        external_retrieval_model: {
+          top_k: externalRetrievalSettings.top_k,
+          score_threshold: externalRetrievalSettings.score_threshold,
+          score_threshold_enabled: externalRetrievalSettings.score_threshold_enabled,
+        },
+      }) as Promise<ExternalKnowledgeBaseHitTestingResponse>,
+    )
+    if (!e) {
+      setExternalHitResult(res)
+      onUpdateList?.()
+    }
+    setLoading(false)
+  }
+
   const retrievalMethod = isEconomy ? RETRIEVE_METHOD.invertedIndex : retrievalConfig.search_method
   const Icon = getIcon(retrievalMethod)
   return (
     <>
       <div className={s.wrapper}>
-        <div className='pt-2 rounded-tl-xl rounded-tr-xl bg-[#EEF4FF]'>
+        <div className='relative pt-2 rounded-tl-xl rounded-tr-xl bg-[#EEF4FF]'>
           <div className="px-4 pb-2 flex justify-between h-8 items-center">
             <span className="text-gray-800 font-semibold text-sm">
               {t('datasetHitTesting.input.title')}
             </span>
-            <Tooltip
-              popupContent={t('dataset.retrieval.changeRetrievalMethod')}
-            >
-              <div
-                onClick={onClickRetrievalMethod}
-                className='flex px-2 h-7 items-center space-x-1 bg-white hover:bg-[#ECE9FE] rounded-md shadow-sm cursor-pointer text-[#6927DA]'
+            {isExternal
+              ? <Button
+                variant='secondary'
+                size='small'
+                onClick={() => setIsSettingsOpen(!isSettingsOpen)}
               >
-                <Icon className='w-3.5 h-3.5'></Icon>
-                <div className='text-xs font-medium'>{t(`dataset.retrieval.${retrievalMethod}.title`)}</div>
-              </div>
-            </Tooltip>
+                <RiEqualizer2Line className='text-components-button-secondary-text w-3.5 h-3.5' />
+                <div className='flex px-[3px] justify-center items-center gap-1'>
+                  <span className='text-components-button-secondary-text system-xs-medium'>{t('datasetHitTesting.settingTitle')}</span>
+                </div>
+              </Button>
+              : <Tooltip
+                popupContent={t('dataset.retrieval.changeRetrievalMethod')}
+              >
+                <div
+                  onClick={onClickRetrievalMethod}
+                  className='flex px-2 h-7 items-center space-x-1 bg-white hover:bg-[#ECE9FE] rounded-md shadow-sm cursor-pointer text-[#6927DA]'
+                >
+                  <Icon className='w-3.5 h-3.5'></Icon>
+                  <div className='text-xs font-medium'>{t(`dataset.retrieval.${retrievalMethod}.title`)}</div>
+                </div>
+              </Tooltip>
+            }
           </div>
+          {
+            isSettingsOpen && (
+              <ModifyExternalRetrievalModal
+                onClose={() => setIsSettingsOpen(false)}
+                onSave={handleSaveExternalRetrievalSettings}
+                initialTopK={externalRetrievalSettings.top_k}
+                initialScoreThreshold={externalRetrievalSettings.score_threshold}
+                initialScoreThresholdEnabled={externalRetrievalSettings.score_threshold_enabled}
+              />
+            )
+          }
           <div className='h-2 rounded-tl-xl rounded-tr-xl bg-white'></div>
         </div>
         <div className='px-4 pb-11'>
@@ -122,7 +184,7 @@ const TextAreaWithButton = ({
 
             <div>
               <Button
-                onClick={onSubmit}
+                onClick={isExternal ? externalRetrievalTestingOnSubmit : onSubmit}
                 variant="primary"
                 loading={loading}
                 disabled={(!text?.length || text?.length > 200)}
@@ -132,7 +194,6 @@ const TextAreaWithButton = ({
             </div>
           </div>
         </div>
-
       </div>
     </>
   )

web/app/components/datasets/rename-modal/index.tsx

@@ -26,6 +26,8 @@ const RenameDatasetModal = ({ show, dataset, onSuccess, onClose }: RenameDataset
   const [loading, setLoading] = useState(false)
   const [name, setName] = useState<string>(dataset.name)
   const [description, setDescription] = useState<string>(dataset.description)
+  const [externalKnowledgeId, setExternalKnowledgeId] = useState<string>(dataset.external_knowledge_info.external_knowledge_id)
+  const [externalKnowledgeApiId, setExternalKnowledgeApiId] = useState<string>(dataset.external_knowledge_info.external_knowledge_api_id)
 
   const onConfirm: MouseEventHandler = async () => {
     if (!name.trim()) {
@@ -34,12 +36,17 @@ const RenameDatasetModal = ({ show, dataset, onSuccess, onClose }: RenameDataset
     }
     try {
       setLoading(true)
+      const body: Partial<DataSet> & { external_knowledge_id?: string; external_knowledge_api_id?: string } = {
+        name,
+        description,
+      }
+      if (externalKnowledgeId && externalKnowledgeApiId) {
+        body.external_knowledge_id = externalKnowledgeId
+        body.external_knowledge_api_id = externalKnowledgeApiId
+      }
       await updateDatasetSetting({
         datasetId: dataset.id,
-        body: {
-          name,
-          description,
-        },
+        body,
       })
       notify({ type: 'success', message: t('common.actionMsg.modifiedSuccessfully') })
       if (onSuccess)

web/app/components/datasets/settings/form/index.tsx

@@ -8,11 +8,14 @@ import { useSWRConfig } from 'swr'
 import { unstable_serialize } from 'swr/infinite'
 import PermissionSelector from '../permission-selector'
 import IndexMethodRadio from '../index-method-radio'
+import RetrievalSettings from '../../external-knowledge-base/create/RetrievalSettings'
 import cn from '@/utils/classnames'
 import RetrievalMethodConfig from '@/app/components/datasets/common/retrieval-method-config'
 import EconomicalRetrievalMethodConfig from '@/app/components/datasets/common/economical-retrieval-method-config'
 import { ToastContext } from '@/app/components/base/toast'
 import Button from '@/app/components/base/button'
+import Divider from '@/app/components/base/divider'
+import { ApiConnectionMod } from '@/app/components/base/icons/src/vender/solid/development'
 import { updateDatasetSetting } from '@/service/datasets'
 import type { DataSetListResponse } from '@/models/datasets'
 import DatasetDetailContext from '@/context/dataset-detail'
@@ -55,6 +58,9 @@ const Form = () => {
   const [name, setName] = useState(currentDataset?.name ?? '')
   const [description, setDescription] = useState(currentDataset?.description ?? '')
   const [permission, setPermission] = useState(currentDataset?.permission)
+  const [topK, setTopK] = useState(currentDataset?.external_retrieval_model.top_k ?? 2)
+  const [scoreThreshold, setScoreThreshold] = useState(currentDataset?.external_retrieval_model.score_threshold ?? 0.5)
+  const [scoreThresholdEnabled, setScoreThresholdEnabled] = useState(currentDataset?.external_retrieval_model.score_threshold_enabled ?? false)
   const [selectedMemberIDs, setSelectedMemberIDs] = useState<string[]>(currentDataset?.partial_member_list || [])
   const [memberList, setMemberList] = useState<Member[]>([])
   const [indexMethod, setIndexMethod] = useState(currentDataset?.indexing_technique)
@@ -85,6 +91,15 @@ const Form = () => {
       setMemberList(accounts)
   }
 
+  const handleSettingsChange = (data: { top_k?: number; score_threshold?: number; score_threshold_enabled?: boolean }) => {
+    if (data.top_k !== undefined)
+      setTopK(data.top_k)
+    if (data.score_threshold !== undefined)
+      setScoreThreshold(data.score_threshold)
+    if (data.score_threshold_enabled !== undefined)
+      setScoreThresholdEnabled(data.score_threshold_enabled)
+  }
+
   useMount(() => {
     getMembers()
   })
@@ -126,10 +141,17 @@ const Form = () => {
           description,
           permission,
           indexing_technique: indexMethod,
+          external_retrieval_model: {
+            top_k: topK,
+            score_threshold: scoreThreshold,
+            score_threshold_enabled: scoreThresholdEnabled,
+          },
           retrieval_model: {
             ...postRetrievalConfig,
             score_threshold: postRetrievalConfig.score_threshold_enabled ? postRetrievalConfig.score_threshold : 0,
           },
+          external_knowledge_id: currentDataset!.external_knowledge_info.external_knowledge_id,
+          external_knowledge_api_id: currentDataset!.external_knowledge_info.external_knowledge_api_id,
           embedding_model: embeddingModel.model,
           embedding_model_provider: embeddingModel.provider,
         },
@@ -161,7 +183,7 @@ const Form = () => {
     <div className='w-full sm:w-[800px] p-4 sm:px-16 sm:py-6'>
       <div className={rowClass}>
         <div className={labelClass}>
-          <div>{t('datasetSettings.form.name')}</div>
+          <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.name')}</div>
         </div>
         <div className='w-full max-w-[480px]'>
           <input
@@ -174,7 +196,7 @@ const Form = () => {
       </div>
       <div className={rowClass}>
         <div className={labelClass}>
-          <div>{t('datasetSettings.form.desc')}</div>
+          <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.desc')}</div>
         </div>
         <div className='w-full max-w-[480px]'>
           <textarea
@@ -192,7 +214,7 @@ const Form = () => {
       </div>
       <div className={rowClass}>
         <div className={labelClass}>
-          <div>{t('datasetSettings.form.permissions')}</div>
+          <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.permissions')}</div>
         </div>
         <div className='w-full sm:w-[480px]'>
           <PermissionSelector
@@ -210,7 +232,7 @@ const Form = () => {
           <div className='w-full h-0 border-b-[0.5px] border-b-gray-200 my-2' />
           <div className={rowClass}>
             <div className={labelClass}>
-              <div>{t('datasetSettings.form.indexMethod')}</div>
+              <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.indexMethod')}</div>
             </div>
             <div className='w-full sm:w-[480px]'>
               <IndexMethodRadio
@@ -225,7 +247,7 @@ const Form = () => {
       {indexMethod === 'high_quality' && (
         <div className={rowClass}>
           <div className={labelClass}>
-            <div>{t('datasetSettings.form.embeddingModel')}</div>
+            <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.embeddingModel')}</div>
           </div>
           <div className='w-[480px]'>
             <ModelSelector
@@ -240,32 +262,76 @@ const Form = () => {
         </div>
       )}
       {/* Retrieval Method Config */}
-      <div className={rowClass}>
-        <div className={labelClass}>
-          <div>
-            <div>{t('datasetSettings.form.retrievalSetting.title')}</div>
-            <div className='leading-[18px] text-xs font-normal text-gray-500'>
-              <a target='_blank' rel='noopener noreferrer' href='https://docs.dify.ai/guides/knowledge-base/create-knowledge-and-upload-documents#id-4-retrieval-settings' className='text-[#155eef]'>{t('datasetSettings.form.retrievalSetting.learnMore')}</a>
-              {t('datasetSettings.form.retrievalSetting.description')}
+      {currentDataset?.provider === 'external'
+        ? <>
+          <div className={rowClass}><Divider/></div>
+          <div className={rowClass}>
+            <div className={labelClass}>
+              <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.retrievalSetting.title')}</div>
             </div>
+            <RetrievalSettings
+              topK={topK}
+              scoreThreshold={scoreThreshold}
+              scoreThresholdEnabled={scoreThresholdEnabled}
+              onChange={handleSettingsChange}
+              isInRetrievalSetting={true}
+            />
+          </div>
+          <div className={rowClass}><Divider/></div>
+          <div className={rowClass}>
+            <div className={labelClass}>
+              <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.externalKnowledgeAPI')}</div>
+            </div>
+            <div className='w-full max-w-[480px]'>
+              <div className='flex h-full px-3 py-2 items-center gap-1 rounded-lg bg-components-input-bg-normal'>
+                <ApiConnectionMod className='w-4 h-4 text-text-secondary' />
+                <div className='overflow-hidden text-text-secondary text-ellipsis system-sm-medium'>
+                  {currentDataset?.external_knowledge_info.external_knowledge_api_name}
+                </div>
+                <div className='text-text-tertiary system-xs-regular'>·</div>
+                <div className='text-text-tertiary system-xs-regular'>{currentDataset?.external_knowledge_info.external_knowledge_api_endpoint}</div>
+              </div>
+            </div>
+          </div>
+          <div className={rowClass}>
+            <div className={labelClass}>
+              <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.externalKnowledgeID')}</div>
+            </div>
+            <div className='w-full max-w-[480px]'>
+              <div className='flex h-full px-3 py-2 items-center gap-1 rounded-lg bg-components-input-bg-normal'>
+                <div className='text-text-tertiary system-xs-regular'>{currentDataset?.external_knowledge_info.external_knowledge_id}</div>
+              </div>
+            </div>
+          </div>
+          <div className={rowClass}><Divider/></div>
+        </>
+        : <div className={rowClass}>
+          <div className={labelClass}>
+            <div>
+              <div className='text-text-secondary system-sm-semibold'>{t('datasetSettings.form.retrievalSetting.title')}</div>
+              <div className='leading-[18px] text-xs font-normal text-gray-500'>
+                <a target='_blank' rel='noopener noreferrer' href='https://docs.dify.ai/guides/knowledge-base/create-knowledge-and-upload-documents#id-4-retrieval-settings' className='text-[#155eef]'>{t('datasetSettings.form.retrievalSetting.learnMore')}</a>
+                {t('datasetSettings.form.retrievalSetting.description')}
+              </div>
+            </div>
+          </div>
+          <div className='w-[480px]'>
+            {indexMethod === 'high_quality'
+              ? (
+                <RetrievalMethodConfig
+                  value={retrievalConfig}
+                  onChange={setRetrievalConfig}
+                />
+              )
+              : (
+                <EconomicalRetrievalMethodConfig
+                  value={retrievalConfig}
+                  onChange={setRetrievalConfig}
+                />
+              )}
           </div>
         </div>
-        <div className='w-[480px]'>
-          {indexMethod === 'high_quality'
-            ? (
-              <RetrievalMethodConfig
-                value={retrievalConfig}
-                onChange={setRetrievalConfig}
-              />
-            )
-            : (
-              <EconomicalRetrievalMethodConfig
-                value={retrievalConfig}
-                onChange={setRetrievalConfig}
-              />
-            )}
-        </div>
-      </div>
+      }
       <div className={rowClass}>
         <div className={labelClass} />
         <div className='w-[480px]'>

web/app/components/header/dataset-nav/index.tsx

@@ -51,7 +51,7 @@ const DatasetNav = () => {
       navs={datasetItems.map(dataset => ({
         id: dataset.id,
         name: dataset.name,
-        link: `/datasets/${dataset.id}/documents`,
+        link: dataset.provider === 'external' ? `/datasets/${dataset.id}/hitTesting` : `/datasets/${dataset.id}/documents`,
         icon: dataset.icon,
         icon_background: dataset.icon_background,
       })) as NavItem[]}

web/app/components/workflow/nodes/knowledge-retrieval/components/dataset-item.tsx

@@ -1,13 +1,15 @@
 'use client'
 import type { FC } from 'react'
-import React, { useCallback } from 'react'
+import React, { useCallback, useState } from 'react'
 import { useBoolean } from 'ahooks'
 import {
   RiDeleteBinLine,
   RiEditLine,
 } from '@remixicon/react'
+import { useTranslation } from 'react-i18next'
 import type { DataSet } from '@/models/datasets'
 import { DataSourceType } from '@/models/datasets'
+import ActionButton, { ActionButtonState } from '@/app/components/base/action-button'
 import FileIcon from '@/app/components/base/file-icon'
 import { Folder } from '@/app/components/base/icons/src/vender/solid/files'
 import SettingsModal from '@/app/components/app/configuration/dataset-config/settings-modal'
@@ -30,8 +32,10 @@ const DatasetItem: FC<Props> = ({
   readonly,
 }) => {
   const media = useBreakpoints()
+  const { t } = useTranslation()
   const isMobile = media === MediaType.mobile
   const { formatIndexingTechniqueAndMethod } = useKnowledge()
+  const [isDeleteHovered, setIsDeleteHovered] = useState(false)
 
   const [isShowSettingsModal, {
     setTrue: showSettingsModal,
@@ -43,8 +47,18 @@ const DatasetItem: FC<Props> = ({
     hideSettingsModal()
   }, [hideSettingsModal, onChange])
 
+  const handleRemove = useCallback((e: React.MouseEvent) => {
+    e.stopPropagation()
+    onRemove()
+  }, [onRemove])
+
   return (
-    <div className='flex items-center h-10 justify-between rounded-xl px-2 bg-white border border-gray-200  cursor-pointer group/dataset-item'>
+    <div className={`flex items-center h-10 justify-between rounded-xl px-2 border-[0.5px] 
+      border-components-panel-border-subtle cursor-pointer group/dataset-item 
+      ${isDeleteHovered
+      ? 'bg-state-destructive-hover border-state-destructive-border'
+      : 'bg-components-panel-on-panel-item-bg hover:bg-components-panel-on-panel-item-bg-hover'
+    }`}>
       <div className='w-0 grow flex items-center space-x-1.5'>
         {
           payload.data_source_type === DataSourceType.NOTION
@@ -61,24 +75,36 @@ const DatasetItem: FC<Props> = ({
       </div>
       {!readonly && (
         <div className='hidden group-hover/dataset-item:flex shrink-0 ml-2  items-center space-x-1'>
-          <div
-            className='flex items-center justify-center w-6 h-6 hover:bg-black/5 rounded-md cursor-pointer'
-            onClick={showSettingsModal}
+          <ActionButton
+            onClick={(e) => {
+              e.stopPropagation()
+              showSettingsModal()
+            }}
           >
-            <RiEditLine className='w-4 h-4 text-gray-500' />
-          </div>
-          <div
-            className='flex items-center justify-center w-6 h-6 hover:bg-black/5 rounded-md cursor-pointer'
-            onClick={onRemove}
+            <RiEditLine className='w-4 h-4 flex-shrink-0 text-text-tertiary' />
+          </ActionButton>
+          <ActionButton
+            onClick={handleRemove}
+            state={ActionButtonState.Destructive}
+            onMouseEnter={() => setIsDeleteHovered(true)}
+            onMouseLeave={() => setIsDeleteHovered(false)}
           >
-            <RiDeleteBinLine className='w-4 h-4 text-gray-500' />
-          </div>
+            <RiDeleteBinLine className={`w-4 h-4 flex-shrink-0 ${isDeleteHovered ? 'text-text-destructive' : 'text-text-tertiary'}`} />
+          </ActionButton>
         </div>
       )}
-      <Badge
-        className='group-hover/dataset-item:hidden shrink-0'
-        text={formatIndexingTechniqueAndMethod(payload.indexing_technique, payload.retrieval_model_dict?.search_method)}
-      />
+      {
+        payload.indexing_technique && <Badge
+          className='group-hover/dataset-item:hidden shrink-0'
+          text={formatIndexingTechniqueAndMethod(payload.indexing_technique, payload.retrieval_model_dict?.search_method)}
+        />
+      }
+      {
+        payload.provider === 'external' && <Badge
+          className='group-hover/dataset-item:hidden shrink-0'
+          text={t('dataset.externalTag')}
+        />
+      }
 
       {isShowSettingsModal && (
         <Drawer isOpen={isShowSettingsModal} onClose={hideSettingsModal} footer={null} mask={isMobile} panelClassname='mt-16 mx-2 sm:mr-2 mb-3 !p-0 !max-w-[640px] rounded-xl'>

web/app/components/workflow/nodes/knowledge-retrieval/use-config.ts

@@ -205,9 +205,11 @@ const useConfig = (id: string, payload: KnowledgeRetrievalNodeType) => {
 
   const handleOnDatasetsChange = useCallback((newDatasets: DataSet[]) => {
     const {
-      allEconomic,
       mixtureHighQualityAndEconomic,
+      mixtureInternalAndExternal,
       inconsistentEmbeddingModel,
+      allInternal,
+      allExternal,
     } = getSelectedDatasetsMode(newDatasets)
     const newInputs = produce(inputs, (draft) => {
       draft.dataset_ids = newDatasets.map(d => d.id)
@@ -220,7 +222,11 @@ const useConfig = (id: string, payload: KnowledgeRetrievalNodeType) => {
     setInputs(newInputs)
     setSelectedDatasets(newDatasets)
 
-    if (allEconomic || mixtureHighQualityAndEconomic || inconsistentEmbeddingModel)
+    if (
+      (allInternal && (mixtureHighQualityAndEconomic || inconsistentEmbeddingModel))
+      || mixtureInternalAndExternal
+      || (allExternal && newDatasets.length > 1)
+    )
       setRerankModelOpen(true)
   }, [inputs, setInputs, payload.retrieval_mode])
 

web/app/components/workflow/nodes/knowledge-retrieval/utils.ts

@@ -21,6 +21,9 @@ export const getSelectedDatasetsMode = (datasets: DataSet[]) => {
   let allHighQualityFullTextSearch = true
   let allEconomic = true
   let mixtureHighQualityAndEconomic = true
+  let allExternal = true
+  let allInternal = true
+  let mixtureInternalAndExternal = true
   let inconsistentEmbeddingModel = false
   if (!datasets.length) {
     allHighQuality = false
@@ -29,6 +32,9 @@ export const getSelectedDatasetsMode = (datasets: DataSet[]) => {
     allEconomic = false
     mixtureHighQualityAndEconomic = false
     inconsistentEmbeddingModel = false
+    allExternal = false
+    allInternal = false
+    mixtureInternalAndExternal = false
   }
   datasets.forEach((dataset) => {
     if (dataset.indexing_technique === 'economy') {
@@ -45,8 +51,21 @@ export const getSelectedDatasetsMode = (datasets: DataSet[]) => {
       if (dataset.retrieval_model_dict.search_method !== RETRIEVE_METHOD.fullText)
         allHighQualityFullTextSearch = false
     }
+    if (dataset.provider !== 'external') {
+      allExternal = false
+    }
+    else {
+      allInternal = false
+      allHighQuality = false
+      allHighQualityVectorSearch = false
+      allHighQualityFullTextSearch = false
+      mixtureHighQualityAndEconomic = false
+    }
   })
 
+  if (allExternal || allInternal)
+    mixtureInternalAndExternal = false
+
   if (allHighQuality || allEconomic)
     mixtureHighQualityAndEconomic = false
 
@@ -59,6 +78,9 @@ export const getSelectedDatasetsMode = (datasets: DataSet[]) => {
     allHighQualityFullTextSearch,
     allEconomic,
     mixtureHighQualityAndEconomic,
+    allInternal,
+    allExternal,
+    mixtureInternalAndExternal,
     inconsistentEmbeddingModel,
   } as SelectedDatasetsMode
 }
@@ -70,6 +92,9 @@ export const getMultipleRetrievalConfig = (multipleRetrievalConfig: MultipleRetr
     allHighQualityFullTextSearch,
     allEconomic,
     mixtureHighQualityAndEconomic,
+    allInternal,
+    allExternal,
+    mixtureInternalAndExternal,
     inconsistentEmbeddingModel,
   } = getSelectedDatasetsMode(selectedDatasets)
 
@@ -91,13 +116,13 @@ export const getMultipleRetrievalConfig = (multipleRetrievalConfig: MultipleRetr
     reranking_enable: allEconomic ? reranking_enable : true,
   }
 
-  if (allEconomic || mixtureHighQualityAndEconomic || inconsistentEmbeddingModel)
+  if (allEconomic || mixtureHighQualityAndEconomic || inconsistentEmbeddingModel || allExternal || mixtureInternalAndExternal)
     result.reranking_mode = RerankingModeEnum.RerankingModel
 
-  if (allHighQuality && !inconsistentEmbeddingModel && reranking_mode === undefined)
+  if (allHighQuality && !inconsistentEmbeddingModel && reranking_mode === undefined && allInternal)
     result.reranking_mode = RerankingModeEnum.WeightedScore
 
-  if (allHighQuality && !inconsistentEmbeddingModel && (reranking_mode === RerankingModeEnum.WeightedScore || reranking_mode === undefined) && !weights) {
+  if (allHighQuality && !inconsistentEmbeddingModel && (reranking_mode === RerankingModeEnum.WeightedScore || reranking_mode === undefined) && allInternal && !weights) {
     result.weights = {
       vector_setting: {
         vector_weight: allHighQualityVectorSearch

web/context/external-api-panel-context.tsx

@@ -0,0 +1,28 @@
+'use client'
+
+import React, { createContext, useContext, useState } from 'react'
+
+type ExternalApiPanelContextType = {
+  showExternalApiPanel: boolean
+  setShowExternalApiPanel: (show: boolean) => void
+}
+
+const ExternalApiPanelContext = createContext<ExternalApiPanelContextType | undefined>(undefined)
+
+export const ExternalApiPanelProvider: React.FC<{ children: React.ReactNode }> = ({ children }) => {
+  const [showExternalApiPanel, setShowExternalApiPanel] = useState(false)
+
+  return (
+    <ExternalApiPanelContext.Provider value={{ showExternalApiPanel, setShowExternalApiPanel }}>
+      {children}
+    </ExternalApiPanelContext.Provider>
+  )
+}
+
+export const useExternalApiPanel = () => {
+  const context = useContext(ExternalApiPanelContext)
+  if (context === undefined)
+    throw new Error('useExternalApiPanel must be used within an ExternalApiPanelProvider')
+
+  return context
+}

web/context/external-knowledge-api-context.tsx

@@ -0,0 +1,46 @@
+'use client'
+
+import { createContext, useContext, useMemo } from 'react'
+import type { FC, ReactNode } from 'react'
+import useSWR from 'swr'
+import type { ExternalAPIItem, ExternalAPIListResponse } from '@/models/datasets'
+import { fetchExternalAPIList } from '@/service/datasets'
+
+type ExternalKnowledgeApiContextType = {
+  externalKnowledgeApiList: ExternalAPIItem[]
+  mutateExternalKnowledgeApis: () => Promise<ExternalAPIListResponse | undefined>
+  isLoading: boolean
+}
+
+const ExternalKnowledgeApiContext = createContext<ExternalKnowledgeApiContextType | undefined>(undefined)
+
+export type ExternalKnowledgeApiProviderProps = {
+  children: ReactNode
+}
+
+export const ExternalKnowledgeApiProvider: FC<ExternalKnowledgeApiProviderProps> = ({ children }) => {
+  const { data, mutate: mutateExternalKnowledgeApis, isLoading } = useSWR<ExternalAPIListResponse>(
+    { url: '/datasets/external-knowledge-api' },
+    fetchExternalAPIList,
+  )
+
+  const contextValue = useMemo<ExternalKnowledgeApiContextType>(() => ({
+    externalKnowledgeApiList: data?.data || [],
+    mutateExternalKnowledgeApis,
+    isLoading,
+  }), [data, mutateExternalKnowledgeApis, isLoading])
+
+  return (
+    <ExternalKnowledgeApiContext.Provider value={contextValue}>
+      {children}
+    </ExternalKnowledgeApiContext.Provider>
+  )
+}
+
+export const useExternalKnowledgeApi = () => {
+  const context = useContext(ExternalKnowledgeApiContext)
+  if (context === undefined)
+    throw new Error('useExternalKnowledgeApi must be used within a ExternalKnowledgeApiProvider')
+
+  return context
+}

web/context/modal-context.tsx

@@ -10,6 +10,7 @@ import ModerationSettingModal from '@/app/components/app/configuration/toolbox/m
 import ExternalDataToolModal from '@/app/components/app/configuration/tools/external-data-tool-modal'
 import AnnotationFullModal from '@/app/components/billing/annotation-full/modal'
 import ModelModal from '@/app/components/header/account-setting/model-provider-page/model-modal'
+import ExternalAPIModal from '@/app/components/datasets/external-api/external-api-modal'
 import type {
   ConfigurationMethodEnum,
   CustomConfigurationModelFixedFields,
@@ -23,6 +24,7 @@ import type {
   ApiBasedExtension,
   ExternalDataTool,
 } from '@/models/common'
+import type { CreateExternalAPIReq } from '@/app/components/datasets/external-api/declarations'
 import ModelLoadBalancingEntryModal from '@/app/components/header/account-setting/model-provider-page/model-modal/model-load-balancing-entry-modal'
 import type { ModelLoadBalancingModalProps } from '@/app/components/header/account-setting/model-provider-page/provider-added-card/model-load-balancing-modal'
 import ModelLoadBalancingModal from '@/app/components/header/account-setting/model-provider-page/provider-added-card/model-load-balancing-modal'
@@ -32,7 +34,10 @@ export type ModalState<T> = {
   onCancelCallback?: () => void
   onSaveCallback?: (newPayload: T) => void
   onRemoveCallback?: (newPayload: T) => void
+  onEditCallback?: (newPayload: T) => void
   onValidateBeforeSaveCallback?: (newPayload: T) => boolean
+  isEditMode?: boolean
+  datasetBindings?: { id: string; name: string }[]
 }
 
 export type ModelModalType = {
@@ -52,6 +57,7 @@ export type ModalContextState = {
   setShowPricingModal: () => void
   setShowAnnotationFullModal: () => void
   setShowModelModal: Dispatch<SetStateAction<ModalState<ModelModalType> | null>>
+  setShowExternalKnowledgeAPIModal: Dispatch<SetStateAction<ModalState<CreateExternalAPIReq> | null>>
   setShowModelLoadBalancingModal: Dispatch<SetStateAction<ModelLoadBalancingModalProps | null>>
   setShowModelLoadBalancingEntryModal: Dispatch<SetStateAction<ModalState<LoadBalancingEntryModalType> | null>>
 }
@@ -63,6 +69,7 @@ const ModalContext = createContext<ModalContextState>({
   setShowPricingModal: () => { },
   setShowAnnotationFullModal: () => { },
   setShowModelModal: () => { },
+  setShowExternalKnowledgeAPIModal: () => { },
   setShowModelLoadBalancingModal: () => { },
   setShowModelLoadBalancingEntryModal: () => { },
 })
@@ -86,6 +93,7 @@ export const ModalContextProvider = ({
   const [showModerationSettingModal, setShowModerationSettingModal] = useState<ModalState<ModerationConfig> | null>(null)
   const [showExternalDataToolModal, setShowExternalDataToolModal] = useState<ModalState<ExternalDataTool> | null>(null)
   const [showModelModal, setShowModelModal] = useState<ModalState<ModelModalType> | null>(null)
+  const [showExternalKnowledgeAPIModal, setShowExternalKnowledgeAPIModal] = useState<ModalState<CreateExternalAPIReq> | null>(null)
   const [showModelLoadBalancingModal, setShowModelLoadBalancingModal] = useState<ModelLoadBalancingModalProps | null>(null)
   const [showModelLoadBalancingEntryModal, setShowModelLoadBalancingEntryModal] = useState<ModalState<LoadBalancingEntryModalType> | null>(null)
   const searchParams = useSearchParams()
@@ -122,6 +130,24 @@ export const ModalContextProvider = ({
     setShowModelModal(null)
   }, [showModelModal])
 
+  const handleCancelExternalApiModal = useCallback(() => {
+    setShowExternalKnowledgeAPIModal(null)
+    if (showExternalKnowledgeAPIModal?.onCancelCallback)
+      showExternalKnowledgeAPIModal.onCancelCallback()
+  }, [showExternalKnowledgeAPIModal])
+
+  const handleSaveExternalApiModal = useCallback(async (updatedFormValue: CreateExternalAPIReq) => {
+    if (showExternalKnowledgeAPIModal?.onSaveCallback)
+      showExternalKnowledgeAPIModal.onSaveCallback(updatedFormValue)
+    setShowExternalKnowledgeAPIModal(null)
+  }, [showExternalKnowledgeAPIModal])
+
+  const handleEditExternalApiModal = useCallback(async (updatedFormValue: CreateExternalAPIReq) => {
+    if (showExternalKnowledgeAPIModal?.onEditCallback)
+      showExternalKnowledgeAPIModal.onEditCallback(updatedFormValue)
+    setShowExternalKnowledgeAPIModal(null)
+  }, [showExternalKnowledgeAPIModal])
+
   const handleCancelModelLoadBalancingEntryModal = useCallback(() => {
     showModelLoadBalancingEntryModal?.onCancelCallback?.()
     setShowModelLoadBalancingEntryModal(null)
@@ -173,6 +199,7 @@ export const ModalContextProvider = ({
       setShowPricingModal: () => setShowPricingModal(true),
       setShowAnnotationFullModal: () => setShowAnnotationFullModal(true),
       setShowModelModal,
+      setShowExternalKnowledgeAPIModal,
       setShowModelLoadBalancingModal,
       setShowModelLoadBalancingEntryModal,
     }}>
@@ -245,6 +272,18 @@ export const ModalContextProvider = ({
             />
           )
         }
+        {
+          !!showExternalKnowledgeAPIModal && (
+            <ExternalAPIModal
+              data={showExternalKnowledgeAPIModal.payload}
+              datasetBindings={showExternalKnowledgeAPIModal.datasetBindings ?? []}
+              onSave={handleSaveExternalApiModal}
+              onCancel={handleCancelExternalApiModal}
+              onEdit={handleEditExternalApiModal}
+              isEditMode={showExternalKnowledgeAPIModal.isEditMode ?? false}
+            />
+          )
+        }
         {
           Boolean(showModelLoadBalancingModal) && (
             <ModelLoadBalancingModal {...showModelLoadBalancingModal!} />

web/i18n/en-US/dataset-hit-testing.ts

@@ -1,6 +1,7 @@
 const translation = {
   title: 'Retrieval Testing',
-  desc: 'Test the hitting effect of the Knowledge based on the given query text.',
+  settingTitle: 'Retrieval Setting',
+  desc: 'Test the hitting effect of the Knowledge based on the given query text',
   dateTimeFormat: 'MM/DD/YYYY hh:mm A',
   recents: 'Recents',
   table: {
@@ -23,6 +24,7 @@ const translation = {
   },
   noRecentTip: 'No recent query results here',
   viewChart: 'View VECTOR CHART',
+  viewDetail: 'View Detail',
 }
 
 export default translation

web/i18n/en-US/dataset-settings.ts

@@ -1,13 +1,13 @@
 const translation = {
   title: 'Knowledge settings',
-  desc: 'Here you can modify the properties and working methods of the Knowledge.',
+  desc: 'Here you can modify the properties and retrieval settings of this Knowledge.',
   form: {
     name: 'Knowledge Name',
     namePlaceholder: 'Please enter the Knowledge name',
     nameError: 'Name cannot be empty',
-    desc: 'Knowledge description',
+    desc: 'Knowledge Description',
     descInfo: 'Please write a clear textual description to outline the content of the Knowledge. This description will be used as a basis for matching when selecting from multiple Knowledge for inference.',
-    descPlaceholder: 'Describe what is in this Knowledge. A detailed description allows AI to access the content of the Knowledge in a timely manner. If empty, Dify will use the default hit strategy.',
+    descPlaceholder: 'Describe what\'s in this Knowledge (optional)',
     descWrite: 'Learn how to write a good Knowledge description.',
     permissions: 'Permissions',
     permissionsOnlyMe: 'Only me',
@@ -23,11 +23,14 @@ const translation = {
     embeddingModelTip: 'Change the embedded model, please go to ',
     embeddingModelTipLink: 'Settings',
     retrievalSetting: {
-      title: 'Retrieval setting',
+      title: 'Retrieval Setting',
       learnMore: 'Learn more',
       description: ' about retrieval method.',
       longDescription: ' about retrieval method, you can change this at any time in the Knowledge settings.',
     },
+    externalKnowledgeAPI: 'External Knowledge API',
+    externalKnowledgeID: 'External Knowledge ID',
+    retrievalSettings: 'Retrieval Settings',
     save: 'Save',
   },
 }

web/i18n/en-US/dataset.ts

@@ -1,9 +1,65 @@
 const translation = {
   knowledge: 'Knowledge',
+  externalTag: 'External',
+  externalAPI: 'External API',
+  externalAPIPanelTitle: 'External Knowledge API',
+  externalKnowledgeId: 'External Knowledge ID',
+  externalKnowledgeName: 'External Knowledge Name',
+  externalKnowledgeDescription: 'Knowledge Description',
+  externalKnowledgeIdPlaceholder: 'Please enter the Knowledge ID',
+  externalKnowledgeNamePlaceholder: 'Please enter the name of the knowledge base',
+  externalKnowledgeDescriptionPlaceholder: 'Describe what\'s in this Knowledge Base (optional)',
+  learnHowToWriteGoodKnowledgeDescription: 'Learn how to write a good knowledge description',
+  externalAPIPanelDescription: 'The external knowledge API is used to connect to a knowledge base outside of Dify and retrieve knowledge from that knowledge base.',
+  externalAPIPanelDocumentation: 'Learn how to create an External Knowledge API',
   documentCount: ' docs',
   wordCount: ' k words',
   appCount: ' linked apps',
   createDataset: 'Create Knowledge',
+  createNewExternalAPI: 'Create a new External Knowledge API',
+  noExternalKnowledge: 'There is no External Knowledge API yet, click here to create',
+  createExternalAPI: 'Add an External Knowledge API',
+  editExternalAPIFormTitle: 'Edit the External Knowledge API',
+  editExternalAPITooltipTitle: 'LINKED KNOWLEDGE',
+  editExternalAPIConfirmWarningContent: {
+    front: 'This External Knowledge API is linked to',
+    end: 'external knowledge, and this modification will be applied to all of them. Are you sure you want to save this change?',
+  },
+  editExternalAPIFormWarning: {
+    front: 'This External API is linked to',
+    end: 'external knowledge',
+  },
+  deleteExternalAPIConfirmWarningContent: {
+    title: {
+      front: 'Delete',
+      end: '?',
+    },
+    content: {
+      front: 'This External Knowledge API is linked to',
+      end: 'external knowledge. Deleting this API will invalidate all of them. Are you sure you want to delete this API?',
+    },
+    noConnectionContent: 'Are you sure to delete this API?',
+  },
+  selectExternalKnowledgeAPI: {
+    placeholder: 'Choose an External Knowledge API',
+  },
+  connectDataset: 'Connect to an External Knowledge Base',
+  connectDatasetIntro: {
+    title: 'How to Connect to an External Knowledge Base',
+    content: {
+      front: 'To connect to an external knowledge base, you need to create an external API first. Please read carefully and refer to',
+      link: 'Learn how to create an external API',
+      end: '. Then find the corresponding knowledge ID and fill it in the form on the left. If all the information is correct, it will automatically jump to the retrieval test in the knowledge base after clicking the connect button.',
+    },
+    learnMore: 'Learn More',
+  },
+  connectHelper: {
+    helper1: 'Connect to external knowledge bases via API and knowledge base ID. Currently, ',
+    helper2: 'only the retrieval functionality is supported',
+    helper3: '. We strongly recommend that you ',
+    helper4: 'read the help documentation',
+    helper5: ' carefully before using this feature.',
+  },
   createDatasetIntro: 'Import your own text data or write data in real-time via Webhook for LLM context enhancement.',
   deleteDatasetConfirmTitle: 'Delete this Knowledge?',
   deleteDatasetConfirmContent:
@@ -22,6 +78,22 @@ const translation = {
   unavailableTip: 'Embedding model is not available, the default embedding model needs to be configured',
   datasets: 'KNOWLEDGE',
   datasetsApi: 'API ACCESS',
+  externalKnowledgeForm: {
+    connect: 'Connect',
+    cancel: 'Cancel',
+  },
+  externalAPIForm: {
+    name: 'Name',
+    endpoint: 'API Endpoint',
+    apiKey: 'API Key',
+    save: 'Save',
+    cancel: 'Cancel',
+    edit: 'Edit',
+    encrypted: {
+      front: 'Your API Token will be encrypted and stored using',
+      end: 'technology.',
+    },
+  },
   retrieval: {
     semantic_search: {
       title: 'Vector Search',
@@ -58,6 +130,8 @@ const translation = {
   defaultRetrievalTip: 'Multi-path retrieval is used by default. Knowledge is retrieved from multiple knowledge bases and then re-ranked.',
   mixtureHighQualityAndEconomicTip: 'The Rerank model is required for mixture of high quality and economical knowledge bases.',
   inconsistentEmbeddingModelTip: 'The Rerank model is required if the Embedding models of the selected knowledge bases are inconsistent.',
+  mixtureInternalAndExternalTip: 'The Rerank model is required for mixture of internal and  external knowledge.',
+  allExternalTip: 'When using external knowledge only, the user can choose whether to enable the Rerank model. If not enabled, retrieved chunks will be sorted based on scores. When the retrieval strategies of different knowledge bases are inconsistent, it will be inaccurate.',
   retrievalSettings: 'Retrieval Setting',
   rerankSettings: 'Rerank Setting',
   weightedScore: {

web/i18n/zh-Hans/dataset-hit-testing.ts

@@ -1,6 +1,7 @@
 const translation = {
   title: '召回测试',
-  desc: '基于给定的查询文本测试知识库的召回效果。',
+  settingTitle: '召回设置',
+  desc: '基于给定的查询文本测试知识库的召回效果',
   dateTimeFormat: 'YYYY-MM-DD HH:mm',
   recents: '最近查询',
   table: {
@@ -23,6 +24,7 @@ const translation = {
   },
   noRecentTip: '最近无查询结果',
   viewChart: '查看向量图表',
+  viewDetail: '查看详情',
 }
 
 export default translation

web/i18n/zh-Hans/dataset-settings.ts

@@ -1,13 +1,13 @@
 const translation = {
   title: '知识库设置',
-  desc: '在这里您可以修改知识库的工作方式以及其它设置。',
+  desc: '在这里，您可以修改此知识库的属性和检索设置',
   form: {
     name: '知识库名称',
     namePlaceholder: '请输入知识库名称',
     nameError: '名称不能为空',
     desc: '知识库描述',
     descInfo: '请写出清楚的文字描述来概述知识库的内容。当从多个知识库中进行选择匹配时，该描述将用作匹配的基础。',
-    descPlaceholder: '描述这个知识库中的内容。详细的描述可以让 AI 及时访问知识库的内容。如果为空，Dify 将使用默认的命中策略。',
+    descPlaceholder: '请描述这个知识库包含的内容（可选）',
     descWrite: '了解如何编写更好的知识库描述。',
     permissions: '可见权限',
     permissionsOnlyMe: '只有我',
@@ -28,6 +28,8 @@ const translation = {
       description: '关于检索方法。',
       longDescription: '关于检索方法，您可以随时在知识库设置中更改此设置。',
     },
+    externalKnowledgeAPI: '外部知识 API',
+    externalKnowledgeID: '外部知识库 ID',
     save: '保存',
   },
 }

web/i18n/zh-Hans/dataset.ts

@@ -1,9 +1,62 @@
 const translation = {
   knowledge: '知识库',
+  externalTag: '外部',
+  externalAPI: '外部 API',
+  externalAPIPanelTitle: '外部知识库 API',
+  externalKnowledgeId: '外部知识库 ID',
+  externalKnowledgeName: '外部知识库名称',
+  externalKnowledgeDescription: '知识库描述',
+  externalKnowledgeIdPlaceholder: '请输入外部知识库 ID',
+  externalKnowledgeNamePlaceholder: '请输入外部知识库名称',
+  externalKnowledgeDescriptionPlaceholder: '描述知识库内容（可选）',
+  learnHowToWriteGoodKnowledgeDescription: '了解如何编写良好的知识库描述',
+  externalAPIPanelDescription: '外部知识库 API 用于连接到 Dify 之外的知识库并从中检索知识。',
+  externalAPIPanelDocumentation: '了解如何创建外部知识库 API',
   documentCount: ' 文档',
   wordCount: ' 千字符',
   appCount: ' 关联应用',
   createDataset: '创建知识库',
+  noExternalKnowledge: '还没有外部知识库 API，点击此处创建',
+  createExternalAPI: '添加外部知识库 API',
+  createNewExternalAPI: '创建新的外部知识库API',
+  editExternalAPIFormTitle: '编辑外部知识库 API',
+  editExternalAPITooltipTitle: '个关联知识库',
+  editExternalAPIConfirmWarningContent: {
+    front: '此外部知识库 API 已链接到',
+    end: '个外部知识库，此修改将应用于所有这些知识库。您确定要保存此更改吗？',
+  },
+  editExternalAPIFormWarning: {
+    front: '此外部 API 已链接到',
+    end: '外部知识库',
+  },
+  deleteExternalAPIConfirmWarningContent: {
+    title: {
+      front: '删除',
+      end: '？',
+    },
+    content: {
+      front: '此外部知识库 API 已链接到',
+      end: '个外部知识库。删除此 API 将使所有这些知识库失效。您确定要删除此 API 吗？',
+    },
+    noConnectionContent: '您确定要删除此 API 吗？',
+  },
+  connectDatasetIntro: {
+    title: '如何连接到外部知识库',
+    content: {
+      front: '要连接到外部知识库，您需要先创建一个外部 API。请仔细阅读并参考',
+      link: '了解如何创建外部 API',
+      end: '。然后找到相应的知识库 ID 并填写在左侧表单中。如果所有信息正确，点击连接按钮后将自动跳转到知识库中的检索测试。',
+    },
+    learnMore: '了解更多',
+  },
+  connectHelper: {
+    helper1: '通过 API 和知识库 ID 连接到外部知识库。目前，',
+    helper2: '仅支持检索功能',
+    helper3: '。我们强烈建议您在使用此功能之前',
+    helper4: '仔细阅读帮助文档',
+    helper5: '。',
+  },
+  connectDataset: '连接外部知识库',
   createDatasetIntro: '导入您自己的文本数据或通过 Webhook 实时写入数据以增强 LLM 的上下文。',
   deleteDatasetConfirmTitle: '要删除知识库吗？',
   deleteDatasetConfirmContent:
@@ -11,6 +64,9 @@ const translation = {
   datasetUsedByApp: '某些应用正在使用该知识库。应用将无法再使用该知识库,所有的提示配置和日志将被永久删除。',
   datasetDeleted: '知识库已删除',
   datasetDeleteFailed: '删除知识库失败',
+  selectExternalKnowledgeAPI: {
+    placeholder: '选择一个外部知识 API',
+  },
   didYouKnow: '你知道吗？',
   intro1: '知识库可以被集成到 Dify 应用中',
   intro2: '作为上下文',
@@ -22,6 +78,22 @@ const translation = {
   unavailableTip: '由于 embedding 模型不可用，需要配置默认 embedding 模型',
   datasets: '知识库',
   datasetsApi: 'API',
+  externalKnowledgeForm: {
+    connect: '连接',
+    cancel: '取消',
+  },
+  externalAPIForm: {
+    name: '名称',
+    endpoint: 'API 端点',
+    apiKey: 'API 密钥',
+    save: '保存',
+    cancel: '取消',
+    edit: '编辑',
+    encrypted: {
+      front: '您的 API Token 将使用',
+      end: '加密并存储。',
+    },
+  },
   retrieval: {
     semantic_search: {
       title: '向量检索',
@@ -58,6 +130,8 @@ const translation = {
   defaultRetrievalTip: '默认情况下使用多路召回。从多个知识库中检索知识，然后重新排序。',
   mixtureHighQualityAndEconomicTip: '混合使用高质量和经济型知识库需要配置 Rerank 模型。',
   inconsistentEmbeddingModelTip: '当所选知识库配置的 Embedding 模型不一致时，需要配置 Rerank 模型。',
+  mixtureInternalAndExternalTip: '混合使用内部和外部知识时需要配置 Rerank 模型。',
+  allExternalTip: '仅使用外部知识时，用户可以选择是否启用 Rerank 模型。如果不启用，检索到的文本块将根据分数排序。当不同知识库的检索策略不一致时，结果可能不准确。',
   retrievalSettings: '召回设置',
   rerankSettings: 'Rerank 设置',
   weightedScore: {

web/i18n/zh-Hant/dataset-hit-testing.ts

@@ -23,6 +23,7 @@ const translation = {
   },
   noRecentTip: '最近無查詢結果',
   viewChart: '查看向量圖表',
+  viewDetail: '查看詳情',
 }
 
 export default translation

web/models/datasets.ts

@@ -25,6 +25,7 @@ export type DataSet = {
   app_count: number
   document_count: number
   word_count: number
+  provider: string
   embedding_model: string
   embedding_model_provider: string
   embedding_available: boolean
@@ -32,6 +33,58 @@ export type DataSet = {
   retrieval_model: RetrievalConfig
   tags: Tag[]
   partial_member_list?: any[]
+  external_knowledge_info: {
+    external_knowledge_id: string
+    external_knowledge_api_id: string
+    external_knowledge_api_name: string
+    external_knowledge_api_endpoint: string
+  }
+  external_retrieval_model: {
+    top_k: number
+    score_threshold: number
+    score_threshold_enabled: boolean
+  }
+}
+
+export type ExternalAPIItem = {
+  id: string
+  tenant_id: string
+  name: string
+  description: string
+  settings: {
+    endpoint: string
+    api_key: string
+  }
+  dataset_bindings: { id: string; name: string }[]
+  created_by: string
+  created_at: string
+}
+
+export type ExternalKnowledgeItem = {
+  id: string
+  name: string
+  description: string | null
+  provider: 'external'
+  permission: DatasetPermission
+  data_source_type: null
+  indexing_technique: null
+  app_count: number
+  document_count: number
+  word_count: number
+  created_by: string
+  created_at: string
+  updated_by: string
+  updated_at: string
+  tags: Tag[]
+}
+
+export type ExternalAPIDeleteResponse = {
+  result: 'success' | 'error'
+}
+
+export type ExternalAPIUsage = {
+  is_using: boolean
+  count: number
 }
 
 export type CustomFile = File & {
@@ -72,6 +125,14 @@ export type DataSetListResponse = {
   total: number
 }
 
+export type ExternalAPIListResponse = {
+  data: ExternalAPIItem[]
+  has_more: boolean
+  limit: number
+  page: number
+  total: number
+}
+
 export type QA = {
   question: string
   answer: string
@@ -384,6 +445,16 @@ export type HitTesting = {
   tsne_position: TsnePosition
 }
 
+export type ExternalKnowledgeBaseHitTesting = {
+  content: string
+  title: string
+  score: number
+  metadata: {
+    'x-amz-bedrock-kb-source-uri': string
+    'x-amz-bedrock-kb-data-source-id': string
+  }
+}
+
 export type Segment = {
   id: string
   document: Document
@@ -424,6 +495,13 @@ export type HitTestingResponse = {
   records: Array<HitTesting>
 }
 
+export type ExternalKnowledgeBaseHitTestingResponse = {
+  query: {
+    content: string
+  }
+  records: Array<ExternalKnowledgeBaseHitTesting>
+}
+
 export type RelatedApp = {
   id: string
   name: string
@@ -461,6 +539,9 @@ export type SelectedDatasetsMode = {
   allHighQualityFullTextSearch: boolean
   allEconomic: boolean
   mixtureHighQualityAndEconomic: boolean
+  allInternal: boolean
+  allExternal: boolean
+  mixtureInternalAndExternal: boolean
   inconsistentEmbeddingModel: boolean
 }
 

web/service/datasets.ts

@@ -8,6 +8,12 @@ import type {
   DocumentDetailResponse,
   DocumentListResponse,
   ErrorDocsResponse,
+  ExternalAPIDeleteResponse,
+  ExternalAPIItem,
+  ExternalAPIListResponse,
+  ExternalAPIUsage,
+  ExternalKnowledgeBaseHitTestingResponse,
+  ExternalKnowledgeItem,
   FileIndexingEstimateResponse,
   HitTestingRecordsResponse,
   HitTestingResponse,
@@ -23,6 +29,8 @@ import type {
   SegmentsResponse,
   createDocumentResponse,
 } from '@/models/datasets'
+import type { CreateKnowledgeBaseReq } from '@/app/components/datasets/external-knowledge-base/create/declarations'
+import type { CreateExternalAPIReq } from '@/app/components/datasets/external-api/declarations.ts'
 import type { CommonResponse, DataSourceNotionWorkspace } from '@/models/common'
 import type {
   ApiKeysListResponse,
@@ -82,6 +90,34 @@ export const deleteDataset: Fetcher<DataSet, string> = (datasetID) => {
   return del<DataSet>(`/datasets/${datasetID}`)
 }
 
+export const fetchExternalAPIList: Fetcher<ExternalAPIListResponse, { url: string }> = ({ url }) => {
+  return get<ExternalAPIListResponse>(url)
+}
+
+export const fetchExternalAPI: Fetcher<ExternalAPIItem, { apiTemplateId: string }> = ({ apiTemplateId }) => {
+  return get<ExternalAPIItem>(`/datasets/external-knowledge-api/${apiTemplateId}`)
+}
+
+export const updateExternalAPI: Fetcher<ExternalAPIItem, { apiTemplateId: string; body: ExternalAPIItem }> = ({ apiTemplateId, body }) => {
+  return patch<ExternalAPIItem>(`/datasets/external-knowledge-api/${apiTemplateId}`, { body })
+}
+
+export const deleteExternalAPI: Fetcher<ExternalAPIDeleteResponse, { apiTemplateId: string }> = ({ apiTemplateId }) => {
+  return del<ExternalAPIDeleteResponse>(`/datasets/external-knowledge-api/${apiTemplateId}`)
+}
+
+export const checkUsageExternalAPI: Fetcher<ExternalAPIUsage, { apiTemplateId: string }> = ({ apiTemplateId }) => {
+  return get<ExternalAPIUsage>(`/datasets/external-knowledge-api/${apiTemplateId}/use-check`)
+}
+
+export const createExternalAPI: Fetcher<ExternalAPIItem, { body: CreateExternalAPIReq }> = ({ body }) => {
+  return post<ExternalAPIItem>('/datasets/external-knowledge-api', { body })
+}
+
+export const createExternalKnowledgeBase: Fetcher<ExternalKnowledgeItem, { body: CreateKnowledgeBaseReq }> = ({ body }) => {
+  return post<ExternalKnowledgeItem>('/datasets/external', { body })
+}
+
 export const fetchDefaultProcessRule: Fetcher<ProcessRuleResponse, { url: string }> = ({ url }) => {
   return get<ProcessRuleResponse>(url)
 }
@@ -209,6 +245,10 @@ export const hitTesting: Fetcher<HitTestingResponse, { datasetId: string; queryT
   return post<HitTestingResponse>(`/datasets/${datasetId}/hit-testing`, { body: { query: queryText, retrieval_model } })
 }
 
+export const externalKnowledgeBaseHitTesting: Fetcher<ExternalKnowledgeBaseHitTestingResponse, { datasetId: string; query: string; external_retrieval_model: { top_k: number; score_threshold: number; score_threshold_enabled: boolean } }> = ({ datasetId, query, external_retrieval_model }) => {
+  return post<ExternalKnowledgeBaseHitTestingResponse>(`/datasets/${datasetId}/external-hit-testing`, { body: { query, external_retrieval_model } })
+}
+
 export const fetchTestingRecords: Fetcher<HitTestingRecordsResponse, { datasetId: string; params: { page: number; limit: number } }> = ({ datasetId, params }) => {
   return get<HitTestingRecordsResponse>(`/datasets/${datasetId}/queries`, { params })
 }