From 49253cbc0479818f04f1d354407843af9f3c9dac Mon Sep 17 00:00:00 2001 From: Wenjie Zhang Date: Wed, 3 Dec 2025 18:28:31 +0800 Subject: [PATCH] =?UTF-8?q?fix(embedding):=20=E4=BF=AE=E5=A4=8Dollama=20mo?= =?UTF-8?q?del=20=E4=B8=8D=E8=B5=B7=E4=BD=9C=E7=94=A8=E7=9A=84=20bug?= =?UTF-8?q?=EF=BC=8C=E5=B9=B6=E4=BF=AE=E5=A4=8Dmodel=20info=20=E7=94=B1?= =?UTF-8?q?=E4=BA=8E=20getattr=20=E4=B8=8D=E7=94=9F=E6=95=88=E7=9A=84=20bu?= =?UTF-8?q?g=20=20#=20361?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - 在 EmbedModelInfo 中添加 model_id 字段 - 修改 get_embedding_config 以优先使用 model_id 选择模型 - 重构 MilvusKB 的嵌入函数获取逻辑 - 添加调试日志并优化错误处理 --- server/routers/knowledge_router.py | 10 ++++--- src/config/static/models.py | 9 ++++++- src/knowledge/implementations/milvus.py | 35 ++++++++++++++++--------- src/knowledge/manager.py | 6 +---- src/knowledge/utils/kb_utils.py | 14 +++++++--- src/models/embed.py | 3 ++- 6 files changed, 50 insertions(+), 27 deletions(-) diff --git a/server/routers/knowledge_router.py b/server/routers/knowledge_router.py index cd6a00f2..e9655fd3 100644 --- a/server/routers/knowledge_router.py +++ b/server/routers/knowledge_router.py @@ -87,10 +87,12 @@ async def create_database( """创建知识库""" logger.debug( f"Create database {database_name} with kb_type {kb_type}, " - f"additional_params {additional_params}, llm_info {llm_info}" + f"additional_params {additional_params}, llm_info {llm_info}, " + f"embed_model_name {embed_model_name}" ) try: additional_params = {**(additional_params or {})} + additional_params["auto_generate_questions"] = False # 默认不生成问题 def normalize_reranker_config(kb: str, params: dict) -> None: reranker_cfg = params.get("reranker_config") @@ -112,12 +114,12 @@ async def create_database( if not isinstance(reranker_cfg, Mapping): raise HTTPException(status_code=400, detail="reranker_config must be an object") - enabled = bool(reranker_cfg.get("enabled", False)) + reranker_enabled = bool(reranker_cfg.get("enabled", False)) model = (reranker_cfg.get("model") or "").strip() recall_top_k = max(1, int(reranker_cfg.get("recall_top_k", 50))) final_top_k = max(1, int(reranker_cfg.get("final_top_k", 10))) - if enabled: + if reranker_enabled: if not model: raise HTTPException(status_code=400, detail="reranker_config.model is required when enabled") if model not in config.reranker_names: @@ -132,7 +134,7 @@ async def create_database( model = model if model in config.reranker_names else "" params["reranker_config"] = { - "enabled": enabled, + "enabled": reranker_enabled, "model": model, "recall_top_k": recall_top_k, "final_top_k": final_top_k, diff --git a/src/config/static/models.py b/src/config/static/models.py index 897ea980..53b5fcbf 100644 --- a/src/config/static/models.py +++ b/src/config/static/models.py @@ -29,7 +29,7 @@ class EmbedModelInfo(BaseModel): dimension: int = Field(..., description="向量维度") base_url: str = Field(..., description="API 基础 URL") api_key: str = Field(..., description="API Key 或环境变量名") - + model_id: str | None = Field(None, description="可选的模型 ID") class RerankerInfo(BaseModel): """重排序模型配置""" @@ -158,42 +158,49 @@ DEFAULT_CHAT_MODEL_PROVIDERS: dict[str, ChatModelProvider] = { DEFAULT_EMBED_MODELS: dict[str, EmbedModelInfo] = { "siliconflow/BAAI/bge-m3": EmbedModelInfo( + model_id="siliconflow/BAAI/bge-m3", name="BAAI/bge-m3", dimension=1024, base_url="https://api.siliconflow.cn/v1/embeddings", api_key="SILICONFLOW_API_KEY", ), "siliconflow/Pro/BAAI/bge-m3": EmbedModelInfo( + model_id="siliconflow/Pro/BAAI/bge-m3", name="Pro/BAAI/bge-m3", dimension=1024, base_url="https://api.siliconflow.cn/v1/embeddings", api_key="SILICONFLOW_API_KEY", ), "siliconflow/Qwen/Qwen3-Embedding-0.6B": EmbedModelInfo( + model_id="siliconflow/Qwen/Qwen3-Embedding-0.6B", name="Qwen/Qwen3-Embedding-0.6B", dimension=1024, base_url="https://api.siliconflow.cn/v1/embeddings", api_key="SILICONFLOW_API_KEY", ), "vllm/Qwen/Qwen3-Embedding-0.6B": EmbedModelInfo( + model_id="vllm/Qwen/Qwen3-Embedding-0.6B", name="Qwen3-Embedding-0.6B", dimension=1024, base_url="http://localhost:8000/v1/embeddings", api_key="no_api_key", ), "ollama/nomic-embed-text": EmbedModelInfo( + model_id="ollama/nomic-embed-text", name="nomic-embed-text", dimension=768, base_url="http://localhost:11434/api/embed", api_key="no_api_key", ), "ollama/bge-m3": EmbedModelInfo( + model_id="ollama/bge-m3", name="bge-m3", dimension=1024, base_url="http://localhost:11434/api/embed", api_key="no_api_key", ), "dashscope/text-embedding-v4": EmbedModelInfo( + model_id="dashscope/text-embedding-v4", name="text-embedding-v4", dimension=1024, base_url="https://dashscope.aliyuncs.com/compatible-mode/v1/embeddings", diff --git a/src/knowledge/implementations/milvus.py b/src/knowledge/implementations/milvus.py index b914b53c..0b22d674 100644 --- a/src/knowledge/implementations/milvus.py +++ b/src/knowledge/implementations/milvus.py @@ -7,6 +7,7 @@ from typing import Any from pymilvus import Collection, CollectionSchema, DataType, FieldSchema, connections, db, utility +from src import config from src.knowledge.base import KnowledgeBase from src.knowledge.indexing import process_file_to_markdown from src.knowledge.utils.kb_utils import ( @@ -91,10 +92,14 @@ class MilvusKB(KnowledgeBase): """创建 Milvus 集合""" logger.info(f"Creating Milvus collection for {db_id}") - if db_id not in self.databases_meta: + if not (metadata := self.databases_meta.get(db_id)): raise ValueError(f"Database {db_id} not found") - embed_info = self.databases_meta[db_id].get("embed_info", {}) + # embed_info = metadata.get("embed_info", {}) + if not (embed_info := metadata.get("embed_info")): + logger.error(f"Embedding info not found for database {db_id}, using default model") + embed_info = config.embed_model_names[config.embed_model] + collection_name = db_id try: @@ -117,8 +122,8 @@ class MilvusKB(KnowledgeBase): except Exception: # 创建新集合 - embedding_dim = getattr(embed_info, "dimension", 1024) if embed_info else 1024 - model_name = getattr(embed_info, "name", "default") if embed_info else "default" + embedding_dim = embed_info.get("dimension", 1024) + model_name = embed_info.get("name", "default") # 定义集合Schema fields = [ @@ -142,7 +147,7 @@ class MilvusKB(KnowledgeBase): index_params = {"metric_type": "COSINE", "index_type": "IVF_FLAT", "params": {"nlist": 1024}} collection.create_index("embedding", index_params) - logger.info(f"Created new Milvus collection: {collection_name}") + logger.info(f"Created new Milvus collection: {collection_name}: {model_name=}, {embedding_dim=}") return collection @@ -154,25 +159,29 @@ class MilvusKB(KnowledgeBase): except Exception as e: logger.warning(f"Failed to load collection into memory: {e}") - def _get_async_embedding_function(self, embed_info: dict): + def _get_async_embedding(self, embed_info: dict): """获取 embedding 函数""" + # 检查是否有 model_id 字段，优先使用 select_embedding_model + if embed_info and "model_id" in embed_info: + from src.models.embed import select_embedding_model + return select_embedding_model(embed_info["model_id"]) + + # 使用原有的逻辑（兼容模式）） config_dict = get_embedding_config(embed_info) - embedding_model = OtherEmbedding( + return OtherEmbedding( model=config_dict.get("model"), base_url=config_dict.get("base_url"), api_key=config_dict.get("api_key"), ) + def _get_async_embedding_function(self, embed_info: dict): + """获取 embedding 函数""" + embedding_model = self._get_async_embedding(embed_info) return partial(embedding_model.abatch_encode, batch_size=40) def _get_embedding_function(self, embed_info: dict): """获取 embedding 函数""" - config_dict = get_embedding_config(embed_info) - embedding_model = OtherEmbedding( - model=config_dict.get("model"), - base_url=config_dict.get("base_url"), - api_key=config_dict.get("api_key"), - ) + embedding_model = self._get_async_embedding(embed_info) return partial(embedding_model.batch_encode, batch_size=40) diff --git a/src/knowledge/manager.py b/src/knowledge/manager.py index 3e912c3f..e52d094f 100644 --- a/src/knowledge/manager.py +++ b/src/knowledge/manager.py @@ -246,16 +246,12 @@ class KnowledgeBaseManager: db_id = db_info["db_id"] async with self._metadata_lock: - # 准备 additional_params，包含 auto_generate_questions - saved_params = kwargs.copy() - saved_params["auto_generate_questions"] = False - self.global_databases_meta[db_id] = { "name": database_name, "description": description, "kb_type": kb_type, "created_at": utc_isoformat(), - "additional_params": saved_params, + "additional_params": kwargs.copy(), } self._save_global_metadata() diff --git a/src/knowledge/utils/kb_utils.py b/src/knowledge/utils/kb_utils.py index 0b06459f..c8d9abcb 100644 --- a/src/knowledge/utils/kb_utils.py +++ b/src/knowledge/utils/kb_utils.py @@ -247,15 +247,23 @@ def get_embedding_config(embed_info: dict) -> dict: try: if embed_info: - # 处理 embed_info 可能是字典或 EmbedModelInfo 对象的情况 - if hasattr(embed_info, "name"): + # 优先检查是否有 model_id 字段 + if "model_id" in embed_info: + from src.models.embed import select_embedding_model + + model = select_embedding_model(embed_info["model_id"]) + config_dict["model"] = model.model + config_dict["api_key"] = model.api_key + config_dict["base_url"] = model.base_url + config_dict["dimension"] = getattr(model, "dimension", 1024) + elif hasattr(embed_info, "name"): # EmbedModelInfo 对象 config_dict["model"] = embed_info.name config_dict["api_key"] = os.getenv(embed_info.api_key) or embed_info.api_key config_dict["base_url"] = embed_info.base_url config_dict["dimension"] = embed_info.dimension else: - # 字典形式 + # 字典形式（保持向后兼容） config_dict["model"] = embed_info["name"] config_dict["api_key"] = os.getenv(embed_info["api_key"]) or embed_info["api_key"] config_dict["base_url"] = embed_info["base_url"] diff --git a/src/models/embed.py b/src/models/embed.py index ebf0ece9..a63bbc58 100644 --- a/src/models/embed.py +++ b/src/models/embed.py @@ -11,7 +11,7 @@ from src.utils import get_docker_safe_url, hashstr, logger class BaseEmbeddingModel(ABC): - def __init__(self, model=None, name=None, dimension=None, url=None, base_url=None, api_key=None): + def __init__(self, model=None, name=None, dimension=None, url=None, base_url=None, api_key=None, model_id=None): """ Args: model: 模型名称，冗余设计，同name @@ -140,6 +140,7 @@ class OllamaEmbedding(BaseEmbeddingModel): payload = {"model": self.model, "input": message} async with httpx.AsyncClient() as client: try: + print(f"\n\n\nOllama Embedding request: {payload}\n\n\n") response = await client.post(self.base_url, json=payload, timeout=60) response.raise_for_status() result = response.json()