feat: enhance agent orchestration, knowledge flow and UI refinements

backend/app/services/settings_service.py

@@ -99,46 +99,55 @@ async def update_scheduler_config(user_id: str, config: dict, db: AsyncSession)
 
 
 async def test_llm_connection(
-    provider: str,
+    provider: str | None,
     model: str,
     base_url: str,
-    api_key: str
+    api_key: str,
 ) -> dict:
     """测试 LLM 连接"""
     try:
+        # base_url-first: provider 可省略
+        from app.services.llm_service import normalize_provider_name
+
+        effective_provider = normalize_provider_name({
+            "provider": provider,
+            "model": model,
+            "base_url": base_url,
+        })
+
         # 根据不同 provider 创建临时 LLM 实例并测试
-        if provider == "openai":
+        if effective_provider in {"openai", "custom", "minimax", "kimi", "qwen"}:
             from langchain_openai import ChatOpenAI
             llm = ChatOpenAI(
                 api_key=api_key,
                 model=model,
                 base_url=base_url or None,
-                timeout=30
+                timeout=30,
             )
-        elif provider == "claude":
+        elif effective_provider == "claude":
             from langchain_anthropic import ChatAnthropic
             llm = ChatAnthropic(
                 api_key=api_key,
                 model=model,
-                timeout=30
+                timeout=30,
             )
-        elif provider == "ollama":
+        elif effective_provider == "ollama":
             from langchain_ollama import ChatOllama
             llm = ChatOllama(
                 base_url=base_url or "http://localhost:11434",
                 model=model,
-                timeout=30
+                timeout=30,
             )
-        elif provider == "deepseek":
+        elif effective_provider == "deepseek":
             from langchain_openai import ChatOpenAI
             llm = ChatOpenAI(
                 api_key=api_key,
                 model=model,
                 base_url=base_url or "https://api.deepseek.com/v1",
-                timeout=30
+                timeout=30,
             )
         else:
-            return {"success": False, "error": f"不支持的 provider: {provider}"}
+            return {"success": False, "error": f"不支持的 endpoint/provider: {effective_provider}"}
 
         # 简单测试调用
         from langchain_core.messages import HumanMessage