diff --git a/apps/scheduler/mcp_agent/plan.py b/apps/scheduler/mcp_agent/plan.py
index 91d293fbfe07fe6c241acef1b4f37345a5e9f853..2d68dd595363764d39c916d78b8d71ece3671e54 100644
--- a/apps/scheduler/mcp_agent/plan.py
+++ b/apps/scheduler/mcp_agent/plan.py
@@ -72,11 +72,12 @@ class MCPPlanner:
 
     @staticmethod
     async def evaluate_goal(
+            goal: str,
             tool_list: list[MCPTool],
             resoning_llm: ReasoningLLM = ReasoningLLM()) -> GoalEvaluationResult:
         """评估用户目标的可行性"""
         # 获取推理结果
-        result = await MCPPlanner._get_reasoning_evaluation(tool_list, resoning_llm)
+        result = await MCPPlanner._get_reasoning_evaluation(goal, tool_list, resoning_llm)
 
         # 解析为结构化数据
         evaluation = await MCPPlanner._parse_evaluation_result(result)
diff --git a/apps/scheduler/scheduler/scheduler.py b/apps/scheduler/scheduler/scheduler.py
index 13eb7ee0b0d81a7a003a0b8fdb913e70f860739d..523587e37deed3258123b1914a5b0fe7563a9130 100644
--- a/apps/scheduler/scheduler/scheduler.py
+++ b/apps/scheduler/scheduler/scheduler.py
@@ -207,9 +207,19 @@ class Scheduler:
         if not app_metadata:
             logger.error("[Scheduler] 未找到Agent应用")
             return
-        llm = await LLMManager.get_llm_by_id(
-            self.task.ids.user_sub, app_metadata.llm_id,
-        )
+        if app_metadata.llm_id == "empty":
+            llm = LLM(
+                _id="empty",
+                user_sub=self.task.ids.user_sub,
+                openai_base_url=Config().get_config().llm.endpoint,
+                openai_api_key=Config().get_config().llm.key,
+                model_name=Config().get_config().llm.model,
+                max_tokens=Config().get_config().llm.max_tokens,
+            )
+        else:
+            llm = await LLMManager.get_llm_by_id(
+                self.task.ids.user_sub, app_metadata.llm_id,
+            )
         if not llm:
             logger.error("[Scheduler] 获取大模型失败")
             await self.queue.close()
diff --git a/apps/schemas/agent.py b/apps/schemas/agent.py
index b52f5e1c3315fb873acddb2bcc4e27937498d561..16e818e4d588d4f25ea4854a7979e2dc5fe90ecb 100644
--- a/apps/schemas/agent.py
+++ b/apps/schemas/agent.py
@@ -17,6 +17,7 @@ class AgentAppMetadata(MetadataBase):
     app_type: AppType = Field(default=AppType.AGENT, description="应用类型", frozen=True)
     published: bool = Field(description="是否发布", default=False)
     history_len: int = Field(description="对话轮次", default=3, le=10)
-    mcp_service: list[str] = Field(default=[], alias="mcpService", description="MCP服务id列表")
+    mcp_service: list[str] = Field(default=[], description="MCP服务id列表")
+    llm_id: str = Field(default="empty", description="大模型ID")
     permission: Permission | None = Field(description="应用权限配置", default=None)
     version: str = Field(description="元数据版本")
diff --git a/apps/schemas/appcenter.py b/apps/schemas/appcenter.py
index a65fffb2db9d74e33145c8faec6995f578ed5e94..2f0549b2cc15772f0a7a0fc5d02853386a8a5945 100644
--- a/apps/schemas/appcenter.py
+++ b/apps/schemas/appcenter.py
@@ -69,6 +69,7 @@ class AppData(BaseModel):
     first_questions: list[str] = Field(
         default=[], alias="recommendedQuestions", description="推荐问题", max_length=3)
     history_len: int = Field(3, alias="dialogRounds", ge=1, le=10, description="对话轮次（1～10）")
+    llm_id: str = Field(default="empty", alias="llmId", description="大模型ID")
     permission: AppPermissionData = Field(
         default_factory=lambda: AppPermissionData(authorizedUsers=None), description="权限配置")
     workflows: list[AppFlowInfo] = Field(default=[], description="工作流信息列表")
diff --git a/apps/schemas/flow.py b/apps/schemas/flow.py
index 2646d04390099fd92988d78c00d1b1471780d6a9..dfffd1f14c779952d9bb761d2e2bccd85528bc0f 100644
--- a/apps/schemas/flow.py
+++ b/apps/schemas/flow.py
@@ -136,6 +136,7 @@ class AppMetadata(MetadataBase):
     published: bool = Field(description="是否发布", default=False)
     links: list[AppLink] = Field(description="相关链接", default=[])
     first_questions: list[str] = Field(description="首次提问", default=[])
+    llm_id: str = Field(description="大模型ID", default="empty")
     history_len: int = Field(description="对话轮次", default=3, le=10)
     permission: Permission | None = Field(description="应用权限配置", default=None)
     flows: list[AppFlow] = Field(description="Flow列表", default=[])
diff --git a/apps/services/appcenter.py b/apps/services/appcenter.py
index e256ab55ab13bf8c5b9e40cb36a3c616a1d39596..559accc2cbf36102e74854bd8f22e72cd9fffcf3 100644
--- a/apps/services/appcenter.py
+++ b/apps/services/appcenter.py
@@ -409,6 +409,7 @@ class AppCenterManager:
             "name": source.name,
             "description": source.description,
             "history_len": source.history_len,
+            "llm_id": source.llm_id,
         }
 
     @staticmethod
diff --git a/deploy/chart/euler_copilot/configs/rag/.env b/deploy/chart/euler_copilot/configs/rag/.env
index 5e83f3f415091b156644b33bef9291a2d99d1931..0708fd0762d1f8c789ee5a4a873fb06880ce1463 100644
--- a/deploy/chart/euler_copilot/configs/rag/.env
+++ b/deploy/chart/euler_copilot/configs/rag/.env
@@ -52,7 +52,7 @@ HALF_KEY3=${halfKey3}
 
 #LLM config
 MODEL_NAME={{ .Values.models.answer.name }}
-OPENAI_API_BASE={{ .Values.models.answer.endpoint }}/v1
+OPENAI_API_BASE={{ .Values.models.answer.endpoint }}
 OPENAI_API_KEY={{ default "" .Values.models.answer.key }}
 MAX_TOKENS={{ default 2048 .Values.models.answer.maxTokens }}