app.py

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. 根路由，用于健康检查
@app.route('/', methods=['GET'])
def root():
    return jsonify({"message": "欢迎使用 GPT-OSS to OpenAI 格式代理API", "status": "ok"})

# 辅助函数：创建一个标准的OpenAI流式数据块
def create_openai_chunk(content, model="gpt-oss-120b"):
    return {
        "id": f"chatcmpl-{str(uuid.uuid4())}",
        "object": "chat.completion.chunk",
        "created": int(time.time()),
        "model": model,
        "choices": [{"index": 0, "delta": {"content": content}, "finish_reason": None}]
    }

# 4. 核心API端点
@app.route('/v1/chat/completions', methods=['POST'])
def chat_completions_proxy():
    try:
        openai_request_data = request.json
        user_prompt = next((m['content'] for m in reversed(openai_request_data.get("messages", [])) if m.get('role') == 'user'), None)
        if not user_prompt: return jsonify({"error": "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

    request_headers = GPT_OSS_HEADERS.copy()
    request_headers['x-show-reasoning'] = 'true' if stream_requested else 'false'
        
    gpt_oss_cookies = {'user_id': str(uuid.uuid4())}
    gpt_oss_payload = {
        "op": "threads.create",
        "params": {"input": {"text": user_prompt, "content": [{"type": "input_text", "text": user_prompt}]}}
    }
    
    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 and line.decode('utf-8').startswith('data: '):
                        try:
                            yield json.loads(line.decode('utf-8')[6:])
                        except json.JSONDecodeError: continue
        except requests.exceptions.RequestException as e:
            raise IOError(f"与后端服务通信失败: {e}")

    if stream_requested:
        def stream_formatter():
            try:
                for gpt_oss_data in _internal_proxy_stream():
                    event_type = gpt_oss_data.get('type')
                    update_type = gpt_oss_data.get('update', {}).get('type')
                    
                    if event_type == 'thread.item_updated' and update_type == 'cot.entry_added':
                        thought = gpt_oss_data['update']['entry']['content']
                        formatted_thought = f"[思考中] {thought}"
                        
                        # 为了兼容性，不逐字发送，而是一次性发送整条思考
                        thought_chunk = create_openai_chunk(formatted_thought)
                        yield f"data: {json.dumps(thought_chunk)}\n\n"
                        # 发送换行符以分隔
                        newline_chunk = create_openai_chunk("\n\n")
                        yield f"data: {json.dumps(newline_chunk)}\n\n"

                    if event_type == 'thread.item_updated' and update_type == 'assistant_message.content_part.text_delta':
                        delta_content = gpt_oss_data['update'].get('delta', '')
                        text_chunk = create_openai_chunk(delta_content)
                        yield f"data: {json.dumps(text_chunk)}\n\n"
                
                yield "data: [DONE]\n\n"
            except IOError as e:
                 yield f"data: {json.dumps({'error': str(e)})}\n\n"

        return Response(stream_formatter(), mimetype='text/event-stream')
    
    else:
        try:
            full_response_content = ""
            for gpt_oss_data in _internal_proxy_stream():
                if gpt_oss_data.get('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', '')
            
            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": {"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)