Update app.py

app.py

@@ -4,23 +4,22 @@ import gradio as gr
 from openai import OpenAI
 
 def encode_image(image_path):
-    """将上传的图片转化为 base64 编码"""
+    """将本地图片文件转换为 Base64 编码"""
     with open(image_path, "rb") as image_file:
         return base64.b64encode(image_file.read()).decode('utf-8')
 
 def predict(message, history):
     api_key = os.getenv("SILICONFLOW_API_KEY", "")
     if not api_key:
-        yield "❌ 未检测到 API Key！请先在 Space Settings -> Secrets 中配置 SILICONFLOW_API_KEY！"
+        yield "❌ 未检测到 API Key！请先在 Space Settings -> Secrets 中添加 SILICONFLOW_API_KEY！"
         return
 
     try:
-        # 初始化 OpenAI 客户端连接硅基流动
         client = OpenAI(
             api_key=api_key.strip(),
             base_url="https://api.siliconflow.cn/v1"
         )
-        
+
         text_content = ""
         files = []
 
@@ -32,7 +31,7 @@ def predict(message, history):
 
         user_content = []
 
-        # 1. 解析图片
+        # 1. 处理上传的图片/文件
         for f in files:
             file_path = f.get("path") if isinstance(f, dict) else f
             if file_path:
@@ -42,17 +41,18 @@ def predict(message, history):
                         "type": "image_url",
                         "image_url": {"url": f"data:image/jpeg;base64,{base64_img}"}
                     })
-                except Exception:
-                    pass
+                except Exception as img_err:
+                    print(f"图片读取失败: {img_err}")
 
-        # 2. 解析文本
+        # 2. 处理输入的文本
         if text_content:
             user_content.append({"type": "text", "text": text_content})
 
         if not user_content:
-            yield "请输入文字或上传照片！"
+            yield "请输入问题或上传图片！"
             return
 
+        # 3. 构造历史对话上下文
         messages = []
         for msg in history:
             role = "user" if msg.get("role") == "user" else "assistant"
@@ -60,16 +60,35 @@ def predict(message, history):
             if isinstance(content_val, dict):
                 content_val = content_val.get("text", "")
             messages.append({"role": role, "content": str(content_val)})
-            
+
         messages.append({"role": "user", "content": user_content})
 
-        # 改用硅基流动目前稳定可用的 Qwen2.5-VL-7B-Instruct 模型
-        response = client.chat.completions.create(
-            model="Qwen/Qwen2.5-VL-7B-Instruct",
-            messages=messages,
-            stream=True
-        )
+        # 4. 尝试请求硅基流动模型（带备选模型降级逻辑）
+        candidate_models = [
+            "Pro/Qwen/Qwen2.5-VL-7B-Instruct",
+            "Qwen/Qwen2-VL-72B-Instruct",
+            "deepseek-ai/deepseek-vl2"
+        ]
+
+        response = None
+        last_exception = None
+
+        for model_name in candidate_models:
+            try:
+                response = client.chat.completions.create(
+                    model=model_name,
+                    messages=messages,
+                    stream=True
+                )
+                break  # 调用成功，跳出循环
+            except Exception as err:
+                last_exception = err
+                continue
+
+        if response is None:
+            raise last_exception
 
+        # 5. 流式输出结果
         partial_text = ""
         for chunk in response:
             if chunk.choices and chunk.choices[0].delta.content:
@@ -79,13 +98,13 @@ def predict(message, history):
     except Exception as e:
         yield f"❌ 请求失败: {str(e)}"
 
-# 构建界面
+# 创建 Gradio 界面
 demo = gr.ChatInterface(
     fn=predict,
     title="🌱 AI 植物与通用多模态识别助手",
-    description="基于 SiliconFlow (Qwen2.5-VL-7B) 驱动，支持照片识别植物与流畅对话",
+    description="基于 SiliconFlow 多模态大模型驱动，支持照片识别与流畅对话",
     multimodal=True
 )
 
 if __name__ == "__main__":
-    demo.launch(ssr_mode=False)
+    demo.launch(ssr_mode=False)  # 关闭 SSR 避免 Gradio 6 探测报错