diff --git a/.env.example b/.env.example
index 65b8bf1..d9f3bc6 100644
--- a/.env.example
+++ b/.env.example
@@ -61,7 +61,7 @@ WORK_ORDER_APP_ID=alarm-system
 WORK_ORDER_APP_SECRET=your-app-secret
 WORK_ORDER_TENANT_ID=1
 
-# ===== 交互Agent配置 =====
+# ===== 交互Agent配置（统一使用 VLM 多模态模型） =====
 AGENT_ENABLED=false
-AGENT_LLM_MODEL=qwen-plus    # 文本对话模型（复用 DASHSCOPE_API_KEY）
-AGENT_LLM_TIMEOUT=15
+AGENT_VLM_MODEL=qwen3-vl-flash-2026-01-22    # 多模态模型（文字+图片）
+AGENT_VLM_TIMEOUT=15
diff --git a/app/config.py b/app/config.py
index 7c6be34..dce8945 100644
--- a/app/config.py
+++ b/app/config.py
@@ -67,11 +67,11 @@ class WeChatConfig:
 
 @dataclass
 class AgentConfig:
-    """交互Agent配置"""
-    llm_api_key: str = ""
-    llm_base_url: str = "https://dashscope.aliyuncs.com/compatible-mode/v1"
-    llm_model: str = "qwen-plus"
-    llm_timeout: int = 15
+    """交互Agent配置（统一使用 VLM 多模态模型）"""
+    vlm_api_key: str = ""
+    vlm_base_url: str = "https://dashscope.aliyuncs.com/compatible-mode/v1"
+    vlm_model: str = "qwen3-vl-flash-2026-01-22"
+    vlm_timeout: int = 15
     enabled: bool = False
 
 
@@ -186,10 +186,10 @@ def load_settings() -> Settings:
             group_chat_id=os.getenv("WECHAT_GROUP_CHAT_ID", ""),
         ),
         agent=AgentConfig(
-            llm_api_key=os.getenv("DASHSCOPE_API_KEY", ""),
-            llm_base_url=os.getenv("AGENT_LLM_BASE_URL", "https://dashscope.aliyuncs.com/compatible-mode/v1"),
-            llm_model=os.getenv("AGENT_LLM_MODEL", "qwen-plus"),
-            llm_timeout=int(os.getenv("AGENT_LLM_TIMEOUT", "15")),
+            vlm_api_key=os.getenv("DASHSCOPE_API_KEY", ""),
+            vlm_base_url=os.getenv("AGENT_VLM_BASE_URL", "https://dashscope.aliyuncs.com/compatible-mode/v1"),
+            vlm_model=os.getenv("AGENT_VLM_MODEL", os.getenv("VLM_MODEL", "qwen3-vl-flash-2026-01-22")),
+            vlm_timeout=int(os.getenv("AGENT_VLM_TIMEOUT", "15")),
             enabled=os.getenv("AGENT_ENABLED", "false").lower() == "true",
         ),
         work_order=WorkOrderConfig(