Update app.py

app.py

@@ -5,15 +5,16 @@ import fitz
 import docx
 from huggingface_hub import InferenceClient
 
-# 1. 强制显式调用 Token
+# 1. 引擎配置与强制授权
+# 关键修复：headers 在此处传入
 hf_token = os.getenv("HF_TOKEN")
 model_id = "MiniMaxAI/MiniMax-M2.5"
 
-# 如果没有 Token，界面会直接提示
-if not hf_token:
-    print("❌ 警告：未检测到 HF_TOKEN，请在 Settings -> Secrets 中添加")
-
-client = InferenceClient(model_id, token=hf_token)
+client = InferenceClient(
+    model=model_id, 
+    token=hf_token,
+    headers={"Authorization": f"Bearer {hf_token}"} if hf_token else None
+)
 
 def extract_text(file_path):
     if not file_path: return ""
@@ -21,7 +22,9 @@ def extract_text(file_path):
         ext = str(file_path).lower()
         if ext.endswith(".pdf"):
             doc = fitz.open(file_path)
-            return "".join([page.get_text() for page in doc])[:15000]
+            text = "".join([page.get_text() for page in doc])
+            doc.close()
+            return text[:15000]
         elif ext.endswith(".docx"):
             doc = docx.Document(file_path)
             return "\n".join([p.text for p in doc.paragraphs])[:15000]
@@ -34,7 +37,7 @@ def extract_text(file_path):
 
 @spaces.GPU(duration=60)
 def chat_fn(message, history):
-    # 彻底压扁历史记录为标准 API 格式
+    # 彻底压扁历史记录
     clean_messages = []
     for turn in history:
         clean_messages.append({"role": turn["role"], "content": str(turn["content"])})
@@ -47,41 +50,42 @@ def chat_fn(message, history):
         file_path = f.get("path") if isinstance(f, dict) else f
         context = extract_text(file_path)
         if context:
-            prompt_text = f"【参考文档】:\n{context}\n\n【指令说明】: 请基于以上内容回答：{prompt_text}"
+            prompt_text = f"【参考文档】:\n{context}\n\n【用户指令】: {prompt_text}"
         
     clean_messages.append({"role": "user", "content": prompt_text})
 
     response = ""
     try:
-        # 增加 headers 确保 Token 穿透
-        for msg in client.chat_completion(
+        # 修复：移除 chat_completion 里的 headers 参数
+        stream = client.chat_completion(
             clean_messages, 
             max_tokens=2048, 
             stream=True, 
-            temperature=0.7,
-            headers={"Authorization": f"Bearer {hf_token}"} if hf_token else None
-        ):
-            if msg.choices[0].delta.content:
-                response += msg.choices[0].delta.content
-                yield response
+            temperature=0.7
+        )
+        
+        for msg in stream:
+            if hasattr(msg, 'choices') and len(msg.choices) > 0:
+                delta = msg.choices[0].delta
+                if hasattr(delta, 'content') and delta.content:
+                    response += delta.content
+                    yield response
         
         if response:
             yield response + "\n\n---\n> **OpenIPOS MegaNode | MiniMax-M2.5 极速逻辑节点**"
             
     except Exception as e:
-        # 如果是权限问题，这里会给出明确提示
-        err_msg = str(e)
-        if "401" in err_msg or "token" in err_msg.lower():
-            yield "❌ 授权失败：请检查 Space Settings 中的 HF_TOKEN 是否正确配置。"
-        else:
-            yield f"⚠️ 调度波动: {err_msg}。请尝试刷新重试。"
+        yield f"⚠️ 算力调度提示: {str(e)}。由于万亿级模型访问量大，若持续报错请检查 Token 权限。"
 
+# 3. 极简黑金 UI
 with gr.Blocks(fill_height=True) as demo:
     gr.HTML("""
         <div style="display: flex; align-items: center; gap: 20px; padding: 10px;">
-            <img src="https://s3.bmp.ovh/2026/03/30/rIAGKg0O.png" style="height: 60px;">
-            <div><h1 style="margin: 0; font-size: 24px;">OpenIPOS MegaNode</h1>
-            <p style="margin: 0; color: #666;">中文精品人工智能算力出口 (MiniMax-M2.5)</p></div>
+            <img src="https://s3.bmp.ovh/2026/03/30/rIAGKg0O.png" style="height: 60px; width: auto;">
+            <div>
+                <h1 style="margin: 0; font-size: 24px;">OpenIPOS MegaNode</h1>
+                <p style="margin: 0; color: #666;">中文精品人工智能算力出口 (MiniMax-M2.5)</p>
+            </div>
         </div>
     """)
     gr.ChatInterface(chat_fn, multimodal=True)