Update app.py

app.py

@@ -9,73 +9,72 @@ from huggingface_hub import InferenceClient
 hf_token = os.getenv("HF_TOKEN")
 client = InferenceClient("deepseek-ai/DeepSeek-R1-Distill-Qwen-32B", token=hf_token)
 
-# 2. 全能文档解析引擎
+# 2. 全能文档解析引擎 (增加格式容错)
 def extract_text_from_file(file_path):
     try:
-        if file_path.lower().endswith(".pdf"):
+        suffix = file_path.lower()
+        if suffix.endswith(".pdf"):
             doc = fitz.open(file_path)
             return "".join([page.get_text() for page in doc])
-        elif file_path.lower().endswith(".docx"):
+        elif suffix.endswith(".docx"):
             doc = docx.Document(file_path)
             return "\n".join([para.text for para in doc.paragraphs])
+        elif suffix.endswith(".doc"):
+            return "⚠️ 提示：系统暂不支持旧版 .doc 格式，请将其另存为 .docx 或 .pdf 后再次上传。"
         else:
             with open(file_path, "r", encoding="utf-8") as f:
                 return f.read()
     except Exception as e:
-        return f"[文档读取失败: {str(e)}]"
+        return f"[解析失败: {str(e)}]"
 
 @spaces.GPU(duration=120)
 def chat_fn(message, history):
-    # 处理当前输入
     files = message.get("files", [])
     current_text = message.get("text", "")
     
     if files:
         file_path = files[0] if isinstance(files[0], str) else files[0].path
         doc_content = extract_text_from_file(file_path)
-        # 截取前 40,000 字以防溢出
-        user_payload = f"【参考文档内容】:\n{doc_content[:40000]}\n\n【用户指令】: {current_text}"
+        # 严格控制输入长度：截取前 30,000 字以换取更高的翻译稳定性
+        user_payload = f"【文档内容参考】:\n{doc_content[:30000]}\n\n【用户指令】: {current_text}"
     else:
         user_payload = current_text
 
-    # 3. 构建严谨的消息列表 (适配 Gradio 6.0)
+    # 严谨的消息过滤
     messages = []
     for turn in history:
-        # 只保留 role 和 content，过滤掉其他干扰字段
-        if isinstance(turn, dict) and "role" in turn and "content" in turn:
-            # 过滤掉空的 content 或非字符串内容
-            if turn["content"] and isinstance(turn["content"], str):
-                messages.append({"role": turn["role"], "content": turn["content"]})
+        if isinstance(turn, dict) and turn.get("role") and turn.get("content"):
+            messages.append({"role": turn["role"], "content": turn["content"]})
     
-    # 加入当前请求
     messages.append({"role": "user", "content": user_payload})
     
     response = ""
     try:
-        # 这里的 client.chat_completion 对 payload 非常挑剔
+        # 增加 stream 稳定性处理
         for msg in client.chat_completion(
             messages, 
             max_tokens=4096, 
             stream=True, 
             temperature=0.6
         ):
-            if len(msg.choices) > 0:
+            if msg.choices:
                 token = msg.choices[0].delta.content
                 if token:
                     response += token
                     yield response
         
-        # 注入品牌签名
+        # 正常结束注入指纹
         response += "\n\n--- \n> **Powered by OpenIPOS Global 算力节点 | 访问 openipos.com 获取企业级私有部署**"
         yield response
         
     except Exception as e:
-        yield f"⚠️ OpenIPOS 调度异常 (API 拒绝请求): {str(e)}\n提示：请尝试清空对话历史后再试。"
+        # 针对超时的友好提示
+        yield f"{response}\n\n⚠️ OpenIPOS 调度异常 (可能是文档过长导致超时): {str(e)}\n提示：请尝试清空对话或分段上传文档。"
 
-# 4. 构建 UI
+# 3. 极简 UI
 with gr.Blocks(fill_height=True) as demo:
     gr.Markdown("# 🚀 OpenIPOS/DeepSeek-R1-MegaNode-64K")
-    gr.Markdown("### 适配 Gradio 6.0 的高稳定性通用算力节点")
+    gr.Markdown("### 适配 Gradio 6.0 | 支持 PDF/DOCX/TXT")
     
     gr.ChatInterface(chat_fn, multimodal=True)