feat: 添加同步获取embedding向量和生成响应的方法

src/chat/knowledge/embedding_store.py

@@ -26,7 +26,7 @@ from rich.progress import (
     TextColumn,
 )
 from src.manager.local_store_manager import local_storage
-from src.chat.utils.utils import get_embedding
+from src.chat.utils.utils import get_embedding_sync
 from src.config.config import global_config
 
 
@@ -99,7 +99,7 @@ class EmbeddingStore:
         self.idx2hash = None
 
     def _get_embedding(self, s: str) -> List[float]:
-        return get_embedding(s)
+        return get_embedding_sync(s)
 
     def get_test_file_path(self):
         return EMBEDDING_TEST_FILE

src/chat/knowledge/ie_process.py

@@ -28,7 +28,7 @@ def _extract_json_from_text(text: str) -> dict:
 def _entity_extract(llm_req: LLMRequest, paragraph: str) -> List[str]:
     """对段落进行实体提取，返回提取出的实体列表（JSON格式）"""
     entity_extract_context = prompt_template.build_entity_extract_context(paragraph)
-    response, (reasoning_content, model_name) = llm_req.generate_response_async(entity_extract_context)
+    response, (reasoning_content, model_name) = llm_req.generate_response_sync(entity_extract_context)
 
     entity_extract_result = _extract_json_from_text(response)
     # 尝试load JSON数据
@@ -50,7 +50,7 @@ def _rdf_triple_extract(llm_req: LLMRequest, paragraph: str, entities: list) ->
     rdf_extract_context = prompt_template.build_rdf_triple_extract_context(
         paragraph, entities=json.dumps(entities, ensure_ascii=False)
     )
-    response, (reasoning_content, model_name) = llm_req.generate_response_async(rdf_extract_context)
+    response, (reasoning_content, model_name) = llm_req.generate_response_sync(rdf_extract_context)
 
     entity_extract_result = _extract_json_from_text(response)
     # 尝试load JSON数据

src/chat/knowledge/qa_manager.py

@@ -10,7 +10,7 @@ from .kg_manager import KGManager
 # from .lpmmconfig import global_config
 from .utils.dyn_topk import dyn_select_top_k
 from src.llm_models.utils_model import LLMRequest
-from src.chat.utils.utils import get_embedding
+from src.chat.utils.utils import get_embedding_sync
 from src.config.config import global_config
 
 MAX_KNOWLEDGE_LENGTH = 10000  # 最大知识长度
@@ -36,7 +36,7 @@ class QAManager:
 
         # 生成问题的Embedding
         part_start_time = time.perf_counter()
-        question_embedding = await get_embedding(question)
+        question_embedding = await get_embedding_sync(question)
         if question_embedding is None:
             logger.error("生成问题Embedding失败")
             return None

src/chat/utils/utils.py

@@ -122,6 +122,18 @@ async def get_embedding(text, request_type="embedding"):
     return embedding
 
 
+def get_embedding_sync(text, request_type="embedding"):
+    """获取文本的embedding向量（同步版本）"""
+    # TODO: API-Adapter修改标记
+    llm = LLMRequest(model=global_config.model.embedding, request_type=request_type)
+    try:
+        embedding = llm.get_embedding_sync(text)
+    except Exception as e:
+        logger.error(f"获取embedding失败: {str(e)}")
+        embedding = None
+    return embedding
+
+
 def get_recent_group_speaker(chat_stream_id: str, sender, limit: int = 12) -> list:
     # 获取当前群聊记录内发言的人
     filter_query = {"chat_id": chat_stream_id}