diff --git a/backend/package/yuxi/knowledge/implementations/lightrag.py b/backend/package/yuxi/knowledge/implementations/lightrag.py index af73a4b4..cb8c9fd2 100644 --- a/backend/package/yuxi/knowledge/implementations/lightrag.py +++ b/backend/package/yuxi/knowledge/implementations/lightrag.py @@ -302,7 +302,7 @@ class LightRagKB(KnowledgeBase): ), ) - async def index_file(self, db_id: str, file_id: str, operator_id: str | None = None) -> dict: + async def index_file(self, db_id: str, file_id: str, operator_id: str | None = None, params: dict | None = None) -> dict: """ Index parsed file (Status: INDEXING -> INDEXED/ERROR_INDEXING) @@ -310,6 +310,7 @@ class LightRagKB(KnowledgeBase): db_id: Database ID file_id: File ID operator_id: ID of the user performing the operation + params: Override processing params to apply during indexing (merged on top of stored params) Returns: Updated file metadata @@ -366,14 +367,15 @@ class LightRagKB(KnowledgeBase): markdown_content = await self._read_markdown_from_minio(file_meta["markdown_file"]) file_path = file_meta.get("path") filename = file_meta.get("filename") or file_id - processing_params = resolve_chunk_processing_params( + params = resolve_chunk_processing_params( kb_additional_params=self.databases_meta.get(db_id, {}).get("metadata"), file_processing_params=file_meta.get("processing_params"), + request_params=params, ) - self.files_meta[file_id]["processing_params"] = processing_params + self.files_meta[file_id]["processing_params"] = params await self._save_metadata() - chunks = chunk_markdown(markdown_content, file_id, filename, processing_params) + chunks = chunk_markdown(markdown_content, file_id, filename, params) chunk_input, split_by_character, split_by_character_only = self._prepare_lightrag_insert_payload(chunks) if not chunk_input: chunk_input = markdown_content diff --git a/backend/package/yuxi/knowledge/implementations/milvus.py b/backend/package/yuxi/knowledge/implementations/milvus.py index 56dbd67b..59579ea5 100644 --- a/backend/package/yuxi/knowledge/implementations/milvus.py +++ b/backend/package/yuxi/knowledge/implementations/milvus.py @@ -284,7 +284,7 @@ class MilvusKB(KnowledgeBase): """将文本分割成块""" return chunk_markdown(text, file_id, filename, params) - async def index_file(self, db_id: str, file_id: str, operator_id: str | None = None) -> dict: + async def index_file(self, db_id: str, file_id: str, operator_id: str | None = None, params: dict | None = None) -> dict: """ Index parsed file (Status: INDEXING -> INDEXED/ERROR_INDEXING) @@ -292,6 +292,7 @@ class MilvusKB(KnowledgeBase): db_id: Database ID file_id: File ID operator_id: ID of the user performing the operation + params: Override processing params to apply during indexing (merged on top of stored params) Returns: Updated file metadata @@ -342,10 +343,11 @@ class MilvusKB(KnowledgeBase): if operator_id: self.files_meta[file_id]["updated_by"] = operator_id - # Read processing params inside lock to ensure we get the latest values + # 将传入的 params 作为 request_params，确保用户指定的参数始终覆盖存储的参数 params = resolve_chunk_processing_params( kb_additional_params=self.databases_meta.get(db_id, {}).get("metadata"), file_processing_params=file_meta.get("processing_params"), + request_params=params, ) self.files_meta[file_id]["processing_params"] = params await self._save_metadata() diff --git a/backend/package/yuxi/knowledge/manager.py b/backend/package/yuxi/knowledge/manager.py index b83f276b..77bdc54a 100644 --- a/backend/package/yuxi/knowledge/manager.py +++ b/backend/package/yuxi/knowledge/manager.py @@ -415,10 +415,10 @@ class KnowledgeBaseManager: kb_instance = await self._get_kb_for_database(db_id) return await kb_instance.parse_file(db_id, file_id, operator_id) - async def index_file(self, db_id: str, file_id: str, operator_id: str | None = None) -> dict: + async def index_file(self, db_id: str, file_id: str, operator_id: str | None = None, params: dict | None = None) -> dict: """Index parsed file""" kb_instance = await self._get_kb_for_database(db_id) - return await kb_instance.index_file(db_id, file_id, operator_id) + return await kb_instance.index_file(db_id, file_id, operator_id, params=params) async def update_file_params(self, db_id: str, file_id: str, params: dict, operator_id: str | None = None) -> None: """Update file processing params""" diff --git a/backend/server/routers/knowledge_router.py b/backend/server/routers/knowledge_router.py index b18b7f53..10e891da 100644 --- a/backend/server/routers/knowledge_router.py +++ b/backend/server/routers/knowledge_router.py @@ -442,8 +442,8 @@ async def add_documents( await knowledge_base.update_file_params( db_id, file_id, indexing_params, operator_id=current_user.user_id ) - # 2. 执行入库 - result = await knowledge_base.index_file(db_id, file_id, operator_id=current_user.user_id) + # 2. 执行入库（传入 indexing_params 确保使用的参数与用户设置一致） + result = await knowledge_base.index_file(db_id, file_id, operator_id=current_user.user_id, params=indexing_params) processed_items.append(result) except Exception as index_error: logger.error(f"自动入库失败 {item} (file_id={file_id}): {index_error}") @@ -605,7 +605,7 @@ async def index_documents( await context.set_progress(progress, f"正在入库第 {idx}/{total} 个文档") try: - result = await knowledge_base.index_file(db_id, file_id, operator_id=operator_id) + result = await knowledge_base.index_file(db_id, file_id, operator_id=operator_id, params=params) processed_items.append(result) except Exception as e: logger.error(f"Index failed for {file_id}: {e}")