Merge branch 'feat/2.4.0' into feat/2.4.0-zl

Author: zgqgit

src/backend/bisheng/common/errcode/tool.py

@@ -49,3 +49,8 @@ class ToolSchemaServerError(BaseErrorCode):
 class ToolMcpSchemaError(BaseErrorCode):
     Code: int = 15024
     Msg: str = 'mcpTool configuration parsing failed, please check if the content matchesmcpConfigure Format: {exception}'
+
+
+class ToolMcpStdioError(BaseErrorCode):
+    Code: int = 15025
+    Msg: str = 'mcpTool not support stdio mode'

src/backend/bisheng/core/vectorstore/multi_retriever.py

@@ -41,7 +41,7 @@ async def _aget_relevant_documents(
 
     def parse_doc_with_score(self, docs_with_score: List[tuple[Document, float]]) -> List[Document]:
         """ parse documents with score to documents only """
-        docs_with_score.sort(key=lambda x: x[1])
+        docs_with_score.sort(key=lambda x: x[1], reverse=True)
         if self.finally_k > 0:
             docs_with_score = docs_with_score[:self.finally_k]
         return [doc for doc, _ in docs_with_score]

src/backend/bisheng/knowledge/domain/services/knowledge_service.py

@@ -1,6 +1,7 @@
 import json
 import math
 import os
+import time
 from datetime import datetime
 from typing import Any, Dict, List, Tuple
 from urllib.parse import urlparse
@@ -57,6 +58,7 @@
 )
 from bisheng.knowledge.domain.repositories.interfaces.knowledge_file_repository import KnowledgeFileRepository
 from bisheng.knowledge.domain.repositories.interfaces.knowledge_repository import KnowledgeRepository
+from bisheng.knowledge.domain.schemas.knowledge_rag_schema import Metadata
 from bisheng.knowledge.domain.schemas.knowledge_schema import AddKnowledgeMetadataFieldsReq, \
     UpdateKnowledgeMetadataFieldsReq
 from bisheng.knowledge.domain.services.knowledge_audit_telemetry_service import KnowledgeAuditTelemetryService
@@ -247,6 +249,25 @@ def create_knowledge_base(cls, request, login_user: UserPayload, db_knowledge: K
             vector_client = KnowledgeRag.init_knowledge_milvus_vectorstore_sync(login_user.user_id,
                                                                                 knowledge=db_knowledge,
                                                                                 metadata_schemas=KNOWLEDGE_RAG_METADATA_SCHEMA)
+            # Init Milvus schema avoiding SchemaNotReady concurrently
+            # Need to provide non-nullable fields to satisfy Milvus schema constraints
+            init_ids = vector_client.add_texts(
+                texts=["init_schema"],
+                metadatas=[Metadata(document_id=0,
+                                    knowledge_id=db_knowledge.id,
+                                    abstract="",
+                                    chunk_index=1,
+                                    bbox="",
+                                    page=0,
+                                    upload_time=int(time.time()),
+                                    update_time=int(time.time()),
+                                    uploader="",
+                                    updater="",
+                                    user_metadata={}).model_dump()]
+            )
+            if init_ids:
+                vector_client.delete(ids=init_ids)
+
             es_client = KnowledgeRag.init_knowledge_es_vectorstore_sync(knowledge=db_knowledge,
                                                                         metadata_schemas=KNOWLEDGE_RAG_METADATA_SCHEMA)
             es_client._store._create_index_if_not_exists()
@@ -1260,21 +1281,20 @@ async def copy_knowledge(
         knowldge_dict.pop("id")
         knowldge_dict.pop("create_time")
         knowldge_dict.pop("update_time", None)
-        knowldge_dict["user_id"] = login_user.user_id
-        knowldge_dict["index_name"] = generate_knowledge_index_name()
-        knowldge_dict["collection_name"] = knowldge_dict["index_name"]
+        knowldge_dict.pop("collection_name", None)
+        knowldge_dict.pop("index_name", None)
         knowldge_dict["name"] = f"{knowledge.name} Copy"[:200] if not knowledge_name else knowledge_name[:200]
 
         knowldge_dict["state"] = KnowledgeState.UNPUBLISHED.value
         knowledge_new = Knowledge(**knowldge_dict)
-        target_knowlege = await KnowledgeDao.async_insert_one(knowledge_new)
-        # celery not yetok
+
+        target_knowlege = cls.create_knowledge_base(request, login_user, knowledge_new)
+
         params = {
             "source_knowledge_id": knowledge.id,
             "target_id": target_knowlege.id,
             "login_user_id": login_user.user_id,
         }
-        cls.create_knowledge_hook(request, login_user, target_knowlege)
         file_worker.file_copy_celery.delay(params)
         return target_knowlege
 

src/backend/bisheng/knowledge/rag/pipeline/transformer/preview_cache.py

@@ -24,9 +24,8 @@ def transform_documents(
             for key, val in all_chunk_info.items():
                 one_metadata = val["metadata"]
                 one_metadata.update(self.file_metadata)
-                for k in one_metadata:
-                    if k not in knowledge_metadata_fields:
-                        del one_metadata[k]
+                for k in one_metadata.keys() - knowledge_metadata_fields:
+                    del one_metadata[k]
                 doc = Document(
                     page_content=KnowledgeUtils.aggregate_chunk_metadata(val["text"], one_metadata),
                     metadata=one_metadata,
@@ -35,9 +34,8 @@ def transform_documents(
         else:
             # aggregate chunk
             for doc in documents:
-                for k in list(doc.metadata.keys()):
-                    if k not in knowledge_metadata_fields:
-                        doc.metadata.pop(k)
+                for k in doc.metadata.keys() - knowledge_metadata_fields:
+                    del doc.metadata[k]
                 doc.metadata.update(self.file_metadata)
                 doc.page_content = KnowledgeUtils.aggregate_chunk_metadata(doc.page_content, doc.metadata)
 

src/backend/bisheng/tool/domain/services/tool.py

@@ -15,7 +15,7 @@
 from bisheng.common.errcode.http_error import UnAuthorizedError, NotFoundError
 from bisheng.common.errcode.tool import ToolTypeNotExistsError, ToolTypeRepeatError, ToolTypeNameError, \
     ToolTypeIsPresetError, ToolSchemaDownloadError, ToolSchemaEmptyError, ToolSchemaParseError, ToolSchemaServerError, \
-    ToolMcpSchemaError
+    ToolMcpSchemaError, ToolMcpStdioError
 from bisheng.common.services.config_service import settings
 from bisheng.database.models.group_resource import GroupResourceDao, ResourceTypeEnum, GroupResource
 from bisheng.database.models.role_access import AccessType
@@ -252,7 +252,7 @@ async def parse_mcp_schema(file_content: str) -> GptsToolsTypeRead:
             if not mcp_conf.enable_stdio:
                 client_type, _ = ClientManager.parse_mcp_client_type(tool_type.openapi_schema)
                 if client_type == McpClientType.STDIO.value:
-                    raise ToolMcpSchemaError(msg="mcp not support stdio mode")
+                    raise ToolMcpStdioError()
             # Instantiatemcpservice object, getting a list of tools
             client = await ClientManager.connect_mcp_from_json(result)
 

src/backend/bisheng/worker/knowledge/file_worker.py

@@ -200,7 +200,7 @@ def copy_vector(
 
     es_db = KnowledgeRag.init_knowledge_es_vectorstore_sync(knowledge=target_knowledge)
     if es_db:
-        insert_es(source_data, es_db)
+        insert_es(source_data, es_db, index_name=target_knowledge.index_name)
 
 
 def create_milvus_col_and_es_index(source_konwledge: Knowledge, target_knowledge: Knowledge):
@@ -244,7 +244,7 @@ def insert_milvus(li: List, fields: list, target: Milvus):
     logger.info("copy_done pk_size={}", len(res_list))
 
 
-def insert_es(li: List, target: ElasticsearchStore):
+def insert_es(li: List, target: ElasticsearchStore, index_name: str):
     from elasticsearch.helpers import bulk
 
     res_list = []
@@ -256,15 +256,15 @@ def insert_es(li: List, target: ElasticsearchStore):
         metadata = data
         request = {
             "_op_type": "index",
-            "_index": target.index_name,
+            "_index": index_name,
             "text": text,
             "metadata": metadata,
             "_id": ids[i],
         }
         requests.append(request)
     bulk(target.client, requests)
 
-    target.client.indices.refresh(index=target.index_name)
+    target.client.indices.refresh(index=index_name)
     logger.info("copy_es_done pk_size={}", len(res_list))
 
 

src/backend/bisheng/workstation/domain/services/workstation_service.py

@@ -33,10 +33,10 @@ class WorkStationService(BaseService):
 
     @classmethod
     def update_config(
-        cls,
-        request: Request,
-        login_user: UserPayload,
-        data: WorkstationConfig,
+            cls,
+            request: Request,
+            login_user: UserPayload,
+            data: WorkstationConfig,
     ) -> WorkstationConfig:
         """Update workstation default configuration."""
         config = ConfigDao.get_config(ConfigKeyEnum.WORKSTATION)
@@ -170,11 +170,11 @@ async def update_knowledge_space_config(cls, data: KnowledgeSpaceConfig) -> Know
 
     @classmethod
     def uploadPersonalKnowledge(
-        cls,
-        request: Request,
-        login_user: UserPayload,
-        file_path,
-        background_tasks: BackgroundTasks,
+            cls,
+            request: Request,
+            login_user: UserPayload,
+            file_path,
+            background_tasks: BackgroundTasks,
     ):
         knowledge = KnowledgeDao.get_user_knowledge(
             login_user.user_id,
@@ -204,11 +204,11 @@ def uploadPersonalKnowledge(
 
     @classmethod
     def queryKnowledgeList(
-        cls,
-        request: Request,
-        login_user: UserPayload,
-        page: int,
-        size: int,
+            cls,
+            request: Request,
+            login_user: UserPayload,
+            page: int,
+            size: int,
     ):
         knowledge = KnowledgeDao.get_user_knowledge(
             login_user.user_id,
@@ -228,11 +228,11 @@ def queryKnowledgeList(
 
     @classmethod
     async def queryChunksFromDB(
-        cls,
-        question: str,
-        use_knowledge_param: UseKnowledgeBaseParam,
-        max_token: int,
-        login_user: UserPayload,
+            cls,
+            question: str,
+            use_knowledge_param: UseKnowledgeBaseParam,
+            max_token: int,
+            login_user: UserPayload,
     ) -> tuple[list[Any], None] | tuple[list[Any], Any]:
         """Query relevant knowledge blocks from the database."""
         try:
@@ -301,8 +301,7 @@ async def get_chat_history(cls, chat_id: str, size: int = 4):
         chat_history = []
         messages = await ChatMessageDao.aget_messages_by_chat_id(chat_id, ['question', 'answer'], size)
         for one in messages:
-            extra = json.loads(one.extra) or {}
-            content = extra['prompt'] if 'prompt' in extra else one.message
+            content = one.message
             if one.category == MessageCategory.QUESTION.value:
                 chat_history.append(HumanMessage(content=content))
             elif one.category == MessageCategory.ANSWER.value:

src/frontend/client/public/assets/channel/notop.svg

@@ -203,6 +203,17 @@ export function parseStreamHistoryItem(raw: StreamHistoryItem): ChatMessage {
         } else {
             // User messages: { query, tags? }
             displayText = parsed.query || parsed.text || raw.message;
+            // Re-encode the first tag (if any) into the same `:::tag {...}:::`
+            // prefix the live send path uses, so the user bubble can render
+            // the chip after a history reload.
+            const firstTag = Array.isArray(parsed.tags) ? parsed.tags[0] : null;
+            if (firstTag && typeof firstTag.name === "string") {
+                const tagJson = JSON.stringify({
+                    id: Number(firstTag.id) || 0,
+                    name: firstTag.name,
+                });
+                displayText = `:::tag ${tagJson}:::\n${displayText}`;
+            }
         }
     } catch {
         displayText = raw.message || "";

src/frontend/client/src/api/chatApi.ts

@@ -404,7 +404,7 @@ export function AccountInfoDialog({
                                                 onClick={() => setShowOldPassword(!showOldPassword)}
                                                 className="absolute right-3 top-1/2 -translate-y-1/2 text-[#86909c] hover:text-[#4e5969]"
                                             >
-                                                {showOldPassword ? <EyeOff className="size-4" /> : <Eye className="size-4" />}
+                                                {showOldPassword ? <Eye className="size-4" /> : <EyeOff className="size-4" />}
                                             </button>
                                         </div>
                                     </div>
@@ -427,7 +427,7 @@ export function AccountInfoDialog({
                                                 onClick={() => setShowNewPassword(!showNewPassword)}
                                                 className="absolute right-3 top-1/2 -translate-y-1/2 text-[#86909c] hover:text-[#4e5969]"
                                             >
-                                                {showNewPassword ? <EyeOff className="size-4" /> : <Eye className="size-4" />}
+                                                {showNewPassword ? <Eye className="size-4" /> : <EyeOff className="size-4" />}
                                             </button>
                                         </div>
 
@@ -460,7 +460,7 @@ export function AccountInfoDialog({
                                                 onClick={() => setShowConfirmPassword(!showConfirmPassword)}
                                                 className="absolute right-3 top-1/2 -translate-y-1/2 text-[#86909c] hover:text-[#4e5969]"
                                             >
-                                                {showConfirmPassword ? <EyeOff className="size-4" /> : <Eye className="size-4" />}
+                                                {showConfirmPassword ? <Eye className="size-4" /> : <EyeOff className="size-4" />}
                                             </button>
                                         </div>
                                     </div>