use lite

prompt/llm.py

@@ -13,8 +13,10 @@ SYSTEM_PROMPT = """你是一个网页生成助手。根据用户的需求描述
 5. 在已有代码基础上修改时，返回完整的修改后代码，不要只返回片段
 6. 由于任何外部链接都被屏蔽，使用纯 HTML、CSS 和 JS 实现功能，不要依赖外部库"""
 
+DEFAULT_MODEL = "deepseek-chat"
+
 # Models served by the ARK (Volcengine) endpoint
-ARK_MODELS = {"doubao-seed-2-0-mini-260215"}
+ARK_MODELS = {"doubao-seed-2-0-lite-260215"}
 
 
 def build_messages(history: list[dict]) -> list[dict]:
@@ -35,21 +37,20 @@ def _get_client(model: str) -> tuple[AsyncOpenAI, str]:
             ),
             model,
         )
-    resolved_model = model or settings.LLM_MODEL
     return (
         AsyncOpenAI(
             api_key=settings.LLM_API_KEY,
             base_url=settings.LLM_BASE_URL,
             timeout=120.0,
         ),
-        resolved_model,
+        model or DEFAULT_MODEL,
     )
 
 
 async def stream_chat(history: list[dict], model: str = ""):
     """Stream chat completion from the LLM. Yields content chunks."""
     messages = build_messages(history)
-    client, resolved_model = _get_client(model or settings.LLM_MODEL)
+    client, resolved_model = _get_client(model)
     async with client as c:
         stream = await c.chat.completions.create(
             model=resolved_model,