feat: stabilize rag chat conversations and auto title sync

fastapi_modules/fastapi_leaudit/rag_engine/generator.py

@@ -7,7 +7,7 @@ from typing import AsyncGenerator
 
 import httpx
 
-from fastapi_modules.fastapi_leaudit.rag_engine.config import RAG_CONFIG
+from fastapi_modules.fastapi_leaudit.rag_engine.config import RAG_CONFIG, build_openai_chat_completions_url
 
 DEFAULT_SYSTEM_PROMPT = """你是烟草行业智慧法务小助手，专注于烟草专卖法规、合同管理、行政处罚等相关法律法规。\n\n回答要求：\n- 先用一句话直接回答，再展开详细说明\n- 多个要点用编号列表\n- 关键法条和数字用 **加粗**\n- 分类信息用表格\n- 层级结构用缩进子列表\n- 不要加标题，直接输出正文"""
 
@@ -17,13 +17,14 @@ async def generate_stream(
     context_chunks: list[dict],
     conversation_id: str,
     message_id: str,
+    task_id: str | None = None,
     system_prompt: str = "",
     model: str = "",
     temperature: float | None = None,
     max_tokens: int | None = None,
     dataset_name: str = "",
 ) -> AsyncGenerator[str, None]:
-    task_id = str(uuid.uuid4())
+    task_id = task_id or str(uuid.uuid4())
     created_at = int(time.time())
     _model = model or RAG_CONFIG["LLM_MODEL"]
     _temp = temperature if temperature is not None else RAG_CONFIG["LLM_TEMPERATURE"]
@@ -55,7 +56,7 @@ async def generate_stream(
         async with httpx.AsyncClient(timeout=RAG_CONFIG["LLM_TIMEOUT"]) as client:
             async with client.stream(
                 "POST",
-                f"{RAG_CONFIG['LLM_BASE_URL'].rstrip('/')}" + "/chat/completions",
+                build_openai_chat_completions_url(RAG_CONFIG["LLM_BASE_URL"]),
                 json={
                     "model": _model,
                     "messages": messages,