Update app.py

app.py

@@ -1,32 +1,160 @@
----
-title: My OpenAI Proxy (Stream/Non-Stream)
-emoji: 🚀
-colorFrom: blue
-colorTo: green
-sdk: gradio
-sdk_version: 3.45.0
-python_version: 3.9
-app_file: app.py
-app_port: 7860
----
-
-# GPT-OSS to OpenAI API Proxy
-
-这是一个代理服务，它将标准的OpenAI API请求转换为 `gpt-oss.com` 所需的格式。
-**此版本支持流式和非流式响应。**
-
-## 如何使用
-
-将您的API客户端的 `base_url` 指向本Space的URL，然后使用`/v1/chat/completions`端点。
-
-### 流式请求 (会返回思考过程)
-
-```bash
-curl https://YOUR_SPACE_URL/v1/chat/completions \
-  -H "Content-Type: application/json" \
-  -H "Authorization: Bearer sk-any-key-works" \
-  -d '{
-    "model": "gpt-oss-120b",
-    "messages": [{"role": "user", "content": "你好吗？"}],
-    "stream": true
-  }'
+import requests
+import json
+import uuid
+import time
+from flask import Flask, request, jsonify, Response
+
+# 1. 初始化Flask应用
+app = Flask(__name__)
+
+# 2. gpt-oss.com API的固定配置
+GPT_OSS_API_URL = "https://api.gpt-oss.com/chatkit"
+GPT_OSS_HEADERS = {
+    'authority': 'api.gpt-oss.com',
+    'accept': 'text/event-stream',
+    'content-type': 'application/json',
+    'origin': 'https://gpt-oss.com',
+    'referer': 'https://gpt-oss.com/',
+    'user-agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36',
+    'x-selected-model': 'gpt-oss-120b',
+}
+
+# 3. 新增：根路由，用于健康检查和显示欢迎信息
+@app.route('/', methods=['GET'])
+def root():
+    return jsonify({
+        "message": "欢迎使用 GPT-OSS to OpenAI 格式代理API",
+        "status": "ok",
+        "api_endpoint": "/v1/chat/completions",
+        "features": {
+            "stream": "支持",
+            "non_stream": "支持",
+            "conditional_reasoning": "仅在流式模式下显示思考过程"
+        }
+    })
+
+# 4. 核心：创建OpenAI兼容的API端点
+@app.route('/v1/chat/completions', methods=['POST'])
+def chat_completions_proxy():
+    try:
+        openai_request_data = request.json
+        messages = openai_request_data.get("messages", [])
+        user_prompt = next((m['content'] for m in reversed(messages) if m.get('role') == 'user'), None)
+        
+        if not user_prompt:
+            return jsonify({"error": "在请求中未找到用户消息。"}), 400
+            
+        stream_requested = openai_request_data.get("stream", False)
+    except Exception as e:
+        return jsonify({"error": f"请求格式无效: {e}"}), 400
+
+    # (一) 条件性设置请求头
+    # 创建一个请求头的副本，以避免修改全局常量
+    request_headers = GPT_OSS_HEADERS.copy()
+    if stream_requested:
+        request_headers['x-show-reasoning'] = 'true'
+    else:
+        request_headers['x-show-reasoning'] = 'false'
+        
+    random_user_id = str(uuid.uuid4())
+    gpt_oss_cookies = {'user_id': random_user_id}
+    gpt_oss_payload = {
+        "op": "threads.create",
+        "params": {"input": {"text": user_prompt, "content": [{"type": "input_text", "text": user_prompt}]}}
+    }
+    
+    # (二) 内部生成器，负责获取和解析后端数据
+    def _internal_proxy_stream():
+        try:
+            with requests.post(
+                GPT_OSS_API_URL, headers=request_headers, cookies=gpt_oss_cookies,
+                json=gpt_oss_payload, stream=True, timeout=120
+            ) as response:
+                response.raise_for_status()
+                for line in response.iter_lines():
+                    if line:
+                        line_str = line.decode('utf-8')
+                        if line_str.startswith('data: '):
+                            json_data_str = line_str[6:]
+                            try:
+                                yield json.loads(json_data_str)
+                            except json.JSONDecodeError:
+                                continue
+        except requests.exceptions.RequestException as e:
+            # 在生成器内部抛出错误，以便外部可以捕获
+            raise IOError(f"与后端服务通信失败: {e}")
+
+    # (三) 根据客户端请求，决定如何格式化响应
+    if stream_requested:
+        # 如果客户端请求流式响应
+        def stream_formatter():
+            try:
+                for gpt_oss_data in _internal_proxy_stream():
+                    event_type = gpt_oss_data.get('type')
+                    
+                    # 格式化思考过程
+                    if (event_type == 'thread.item_updated' and 
+                        gpt_oss_data.get('update', {}).get('type') == 'cot.entry_added'):
+                        thought = gpt_oss_data['update']['entry']['content']
+                        # 我们可以选择发送一个自定义的流块，或者忽略它
+                        # 这里我们创建一个包含思考过程的自定义块
+                        reasoning_chunk = { "reasoning": thought }
+                        yield f"data: {json.dumps(reasoning_chunk)}\n\n"
+
+                    # 格式化文本片段
+                    if (event_type == 'thread.item_updated' and 
+                        gpt_oss_data.get('update', {}).get('type') == 'assistant_message.content_part.text_delta'):
+                        delta_content = gpt_oss_data['update'].get('delta', '')
+                        openai_chunk = {
+                            "id": f"chatcmpl-{str(uuid.uuid4())}", "object": "chat.completion.chunk",
+                            "created": int(time.time()), "model": "gpt-oss-120b",
+                            "choices": [{"index": 0, "delta": {"content": delta_content}, "finish_reason": None}]
+                        }
+                        yield f"data: {json.dumps(openai_chunk)}\n\n"
+                
+                yield "data: [DONE]\n\n"
+            except IOError as e:
+                 error_chunk = {"error": str(e)}
+                 yield f"data: {json.dumps(error_chunk)}\n\n"
+
+        return Response(stream_formatter(), mimetype='text/event-stream')
+    
+    else:
+        # 如果客户端请求非流式响应
+        try:
+            full_response_content = ""
+            for gpt_oss_data in _internal_proxy_stream():
+                event_type = gpt_oss_data.get('type')
+                # 只关心最终的文本片段
+                if (event_type == 'thread.item_updated' and 
+                    gpt_oss_data.get('update', {}).get('type') == 'assistant_message.content_part.text_delta'):
+                    full_response_content += gpt_oss_data['update'].get('delta', '')
+            
+            # 构建一个标准的OpenAI非流式JSON响应
+            final_response = {
+                "id": f"chatcmpl-{str(uuid.uuid4())}",
+                "object": "chat.completion",
+                "created": int(time.time()),
+                "model": "gpt-oss-120b",
+                "choices": [
+                    {
+                        "index": 0,
+                        "message": {
+                            "role": "assistant",
+                            "content": full_response_content.strip()
+                        },
+                        "finish_reason": "stop"
+                    }
+                ],
+                "usage": { # 提供一个模拟的usage对象
+                    "prompt_tokens": None, "completion_tokens": None, "total_tokens": None
+                }
+            }
+            return jsonify(final_response)
+        
+        except IOError as e:
+            return jsonify({"error": str(e)}), 500
+
+# 5. 启动应用
+if __name__ == '__main__':
+    app.run(host='0.0.0.0', port=7860)