├── README.md ├── linux └── app.py └── win └── app.py /README.md: -------------------------------------------------------------------------------- 1 | # Chutes2api 2 | ### 对你有用的话麻烦给个stars谢谢 3 | 4 | ## 支持模型 5 | - nvidia/Llama-3.1-405B-Instruct-FP8 6 | - deepseek-ai/DeepSeek-R1 7 | - Qwen/Qwen2.5-72B-Instruct 8 | - Qwen/Qwen2.5-Coder-32B-Instruc 9 | - bytedance-research/UI-TARS-72B-DPO 10 | - OpenGVLab/InternVL2_5-78B 11 | - hugging-quants/Meta-Llama-3.1-70B-Instruct-AWQ-INT4 12 | - NousResearch/Hermes-3-Llama-3.1-8B 13 | - Qwen/QVQ-72B-Preview 14 | - deepseek-ai/DeepSeek-R1-Distill-Qwen-32B 15 | - jondurbin/bagel-8b-v1.0 16 | - unsloth/QwQ-32B-Preview 17 | - Qwen/QwQ-32B-Preview 18 | - jondurbin/airoboros-34b-3.3 19 | - NovaSky-AI/Sky-T1-32B-Preview 20 | - driaforall/Dria-Agent-a-3B 21 | - NousResearch/Nous-Hermes-Llama2-13b 22 | - unsloth/Llama-3.2-1B-Instruct 23 | 24 | 25 | ## 请求路由 26 | - /v1/models 27 | - /v1/chat/completions 28 | - 如果有配置AUTH_TOKEN请求都要带key 29 | 30 | 31 | ## 请求格式 32 | 和 OpenAI 的请求格式相同，支持非流式和流式响应 33 | 34 | ## Docker部署 35 | 36 | ### Linux版本 37 | ```bash 38 | # 拉取 39 | docker pull mtxyt/chutes2api-linux:1.2 40 | 41 | # 运行(不带认证) 42 | docker run -d -p 8805:8805 mtxyt/chutes2api-linux:1.2 43 | 44 | # 运行(带认证) 45 | docker run -d -e AUTH_TOKEN=your_token -p 8805:8805 mtxyt/chutes2api-linux:1.2 46 | ``` 47 | 48 | ### Windows版本 49 | ```bash 50 | # 拉取 51 | docker pull mtxyt/chutes2api:1.5 52 | 53 | # 运行(不带认证) 54 | docker run -d -p 8805:8805 mtxyt/chutes2api:1.5 55 | 56 | # 运行(带认证) 57 | docker run -d -e AUTH_TOKEN=your_token -p 8805:8805 mtxyt/chutes2api:1.5 58 | ``` 59 | 60 | ## Token获取方式 61 | ### 准备步骤 62 | 1. 访问 [chutes.ai](https://chutes.ai) 63 | 2. 随便找一个LLM模型切换到chat聊天界面里 64 | 4. 打开开发者工具(F12)，切换到网络面板 65 | 5. 随便发起一个对话 66 | 6. 在请求中找到 cf_clearance cookie(如果没有,多对话几次回复空,就是触发盾了刷新网页过人机验证再请求就有了) 67 | ![image](https://github.com/user-attachments/assets/9e5423aa-9b4c-4c97-a737-281d3f195884) 68 | 69 | 如果你配置了 AUTH_TOKEN，完整的认证格式为：`AUTH_TOKEN|||cf_clearance` 70 | 71 | 例如: 72 | - AUTH_TOKEN=123456 73 | - cf_clearance=abcdef 74 | - 最终认证格式: `123456|||abcdef` 75 | 76 | ## 注意事项 77 | 1. cf_clearance 有时效性，过期后会自动获取新的 78 | 2. 支持配置和不配置 AUTH_TOKEN 两种方式 79 | 3. 服务启动后可以访问根路径查看服务状态 80 | 81 | ## 更多 82 | 目前提供 Linux 和 Windows 两个版本的镜像，win版本在win系统系统运行时稳定性更好。 83 | 84 | ## 声明 85 | 本项目仅供学习研究使用，请勿用于商业用途。 86 | -------------------------------------------------------------------------------- /linux/app.py: -------------------------------------------------------------------------------- 1 | from flask import Flask, request, Response, stream_with_context, jsonify 2 | import cloudscraper 3 | import json 4 | import uuid 5 | from datetime import datetime, timezone 6 | import time 7 | import os 8 | import logging 9 | 10 | # 配置日志 11 | logging.basicConfig( 12 | level=logging.INFO, 13 | format='%(asctime)s - %(levelname)s - %(message)s' 14 | ) 15 | 16 | app = Flask(__name__) 17 | 18 | # 模型映射字典 19 | MODEL_MAPPING = { 20 | "nvidia/Llama-3.1-405B-Instruct-FP8": "chutes-nvidia-llama-3-1-405b-instruct-fp8", 21 | "deepseek-ai/DeepSeek-R1": "chutes-deepseek-ai-deepseek-r1", 22 | "Qwen/Qwen2.5-72B-Instruct": "chutes-qwen-qwen2-5-72b-instruct", 23 | "Qwen/Qwen2.5-Coder-32B-Instruc": "chutes-qwen-qwen2-5-coder-32b-instruct", 24 | "bytedance-research/UI-TARS-72B-DPO": "chutes-bytedance-research-ui-tars-72b-dpo", 25 | "OpenGVLab/InternVL2_5-78B": "chutes-opengvlab-internvl2-5-78b", 26 | "hugging-quants/Meta-Llama-3.1-70B-Instruct-AWQ-INT4": "chutes-hugging-quants-meta-llama-3-1-70b-instruct-awq-int4", 27 | "NousResearch/Hermes-3-Llama-3.1-8B": "cxmplexbb-nousresearch-hermes-3-llama-3-1-8b", 28 | "Qwen/QVQ-72B-Preview": "chutes-qwen-qvq-72b-preview", 29 | "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B": "chutes-deepseek-ai-deepseek-r1-distill-qwen-32b", 30 | "jondurbin/bagel-8b-v1.0": "chutes-jondurbin-bagel-8b-v1-0", 31 | "unsloth/QwQ-32B-Preview": "cxmplexbb-unsloth-qwq-32b-preview", 32 | "Qwen/QwQ-32B-Preview": "chutes-qwq-32b-preview", 33 | "jondurbin/airoboros-34b-3.3": "chutes-jondurbin-airoboros-34b-3-3", 34 | "NovaSky-AI/Sky-T1-32B-Preview": "chutes-novasky-ai-sky-t1-32b-preview", 35 | "driaforall/Dria-Agent-a-3B": "chutes-driaforall-dria-agent-a-3b", 36 | "NousResearch/Nous-Hermes-Llama2-13b": "cxmplexbb-nousresearch-nous-hermes-llama2-13b", 37 | "unsloth/Llama-3.2-1B-Instruct": "chutes-unsloth-llama-3-2-1b-instruct" 38 | } 39 | 40 | # 存储当前的 cf_clearance 和 key 41 | current_cf_clearance = None 42 | auth_token = os.getenv('AUTH_TOKEN', '') # 从环境变量获取 AUTH_TOKEN 43 | 44 | def check_auth(): 45 | """检查认证""" 46 | request_token = request.headers.get('Authorization', '') 47 | if not auth_token: 48 | return True 49 | return request_token == f"Bearer {auth_token}" 50 | 51 | def get_new_cf_clearance(): 52 | """动态获取新的 cf_clearance""" 53 | try: 54 | temp_scraper = cloudscraper.create_scraper( 55 | browser={ 56 | 'browser': 'firefox', 57 | 'platform': 'windows', 58 | 'mobile': False 59 | }, 60 | delay=10, 61 | allow_brotli=False, 62 | doubleDown=True 63 | ) 64 | 65 | if current_cf_clearance: 66 | if auth_token: 67 | temp_scraper.headers.update({ 68 | "Authorization": f"Bearer {auth_token}|||{current_cf_clearance}" 69 | }) 70 | else: 71 | temp_scraper.headers.update({ 72 | "Authorization": f"Bearer {current_cf_clearance}" 73 | }) 74 | elif auth_token: 75 | temp_scraper.headers.update({ 76 | "Authorization": f"Bearer {auth_token}" 77 | }) 78 | 79 | logging.info("尝试获取新的 cf_clearance") 80 | response = temp_scraper.get('https://chutes.ai') 81 | 82 | if 'cf_clearance' in temp_scraper.cookies: 83 | new_cf = temp_scraper.cookies['cf_clearance'] 84 | logging.info(f"成功获取新的 cf_clearance: {new_cf[:10]}...") 85 | return new_cf 86 | logging.warning("未能获取 cf_clearance") 87 | return None 88 | except Exception as e: 89 | logging.error(f"获取新的 cf_clearance 失败: {str(e)}") 90 | return None 91 | 92 | def create_scraper(cf_clearance=None): 93 | """创建配置好的 scraper""" 94 | global current_cf_clearance 95 | 96 | scraper = cloudscraper.create_scraper( 97 | browser={ 98 | 'browser': 'firefox', 99 | 'platform': 'windows', 100 | 'mobile': False 101 | }, 102 | delay=10, 103 | allow_brotli=False, 104 | doubleDown=True 105 | ) 106 | 107 | # 基础请求头配置 108 | scraper.headers.update({ 109 | "Accept": "text/event-stream", 110 | "Accept-Encoding": "gzip, deflate, br", 111 | "Accept-Language": "en-US,en;q=0.9", 112 | "Cache-Control": "no-cache", 113 | "Content-Type": "application/json", 114 | "Origin": "https://chutes.ai", 115 | "Pragma": "no-cache", 116 | "Referer": "https://chutes.ai/", 117 | "Sec-Ch-Ua": '"Chromium";v="122", "Not(A:Brand";v="24", "Microsoft Edge";v="122"', 118 | "Sec-Ch-Ua-Mobile": "?0", 119 | "Sec-Ch-Ua-Platform": '"Windows"', 120 | "Sec-Fetch-Dest": "empty", 121 | "Sec-Fetch-Mode": "cors", 122 | "Sec-Fetch-Site": "same-origin", 123 | "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36 Edg/122.0.0.0", 124 | "X-Requested-With": "XMLHttpRequest" 125 | }) 126 | 127 | # 使用提供的或当前的 cf_clearance 128 | cf_value = cf_clearance or current_cf_clearance 129 | if cf_value: 130 | scraper.cookies.update({ 131 | "cf_clearance": cf_value 132 | }) 133 | # 构造 Authorization 头 134 | if auth_token: 135 | scraper.headers.update({ 136 | "Authorization": f"Bearer {auth_token}|||{cf_value}" 137 | }) 138 | else: 139 | # 如果没有 auth_token，只使用 cf_clearance 140 | scraper.headers.update({ 141 | "Authorization": f"Bearer {cf_value}" 142 | }) 143 | 144 | return scraper 145 | 146 | def create_chutes_request(openai_request): 147 | """将OpenAI格式请求转换为Chutes格式""" 148 | messages = openai_request['messages'] 149 | message_id = str(uuid.uuid4()) 150 | current_time = datetime.now(timezone.utc).strftime('%Y-%m-%dT%H:%M:%S.%f')[:-3] + 'Z' 151 | 152 | model = openai_request.get('model', 'deepseek-ai/DeepSeek-R1') 153 | chute_name = MODEL_MAPPING.get(model, 'chutes-deepseek-ai-deepseek-r1') 154 | 155 | return { 156 | "messages": [{ 157 | "role": messages[-1]['role'], 158 | "content": messages[-1]['content'], 159 | "id": message_id, 160 | "createdOn": current_time 161 | }], 162 | "model": model, 163 | "chuteName": chute_name 164 | } 165 | 166 | def process_chunk(chunk): 167 | """处理响应数据块""" 168 | try: 169 | if "choices" in chunk and chunk["choices"][0]["delta"].get("content"): 170 | return chunk["choices"][0]["delta"]["content"] 171 | return None 172 | except: 173 | return None 174 | 175 | def process_non_stream_response(response, model): 176 | """处理非流式响应""" 177 | try: 178 | full_content = "" 179 | for line in response.iter_lines(): 180 | if line: 181 | line = line.decode('utf-8') 182 | if line.startswith("data: "): 183 | data = line[6:] 184 | if data == "[DONE]": 185 | break 186 | try: 187 | chunk = json.loads(data) 188 | content = process_chunk(chunk) 189 | if content: 190 | full_content += content 191 | except json.JSONDecodeError: 192 | continue 193 | 194 | if not full_content: 195 | return Response("Empty response from server", status=500) 196 | 197 | return { 198 | "id": str(uuid.uuid4()), 199 | "object": "chat.completion", 200 | "created": int(time.time()), 201 | "model": model, 202 | "choices": [{ 203 | "message": { 204 | "role": "assistant", 205 | "content": full_content 206 | }, 207 | "finish_reason": "stop", 208 | "index": 0 209 | }], 210 | "usage": { 211 | "prompt_tokens": 0, 212 | "completion_tokens": 0, 213 | "total_tokens": 0 214 | } 215 | } 216 | except Exception as e: 217 | logging.error(f"处理非流式响应时出错: {str(e)}") 218 | return Response("Failed to process response", status=500) 219 | 220 | def make_request_with_retry(openai_request, max_retries=3): 221 | """带重试机制的请求函数""" 222 | global current_cf_clearance 223 | 224 | chutes_request = create_chutes_request(openai_request) 225 | last_error = None 226 | 227 | for attempt in range(max_retries): 228 | try: 229 | scraper = create_scraper() 230 | logging.info(f"尝试第 {attempt + 1} 次请求") 231 | 232 | response = scraper.post( 233 | "https://chutes.ai/app/api/chat", 234 | json=chutes_request, 235 | stream=True 236 | ) 237 | 238 | logging.info(f"请求状态码: {response.status_code}") 239 | 240 | # 如果响应成功,返回响应对象 241 | if response.status_code == 200: 242 | return response 243 | 244 | # 如果是 403 错误,尝试获取新的 cf_clearance 245 | if response.status_code == 403: 246 | logging.warning(f"尝试 {attempt + 1}: 获取新的 cf_clearance") 247 | new_cf_clearance = get_new_cf_clearance() 248 | if new_cf_clearance: 249 | current_cf_clearance = new_cf_clearance 250 | continue 251 | 252 | last_error = f"Status code: {response.status_code}, Response: {response.text}" 253 | logging.error(f"请求失败: {last_error}") 254 | 255 | except Exception as e: 256 | last_error = str(e) 257 | logging.error(f"尝试 {attempt + 1} 失败: {last_error}", exc_info=True) 258 | 259 | # 在重试之前等待一段时间 260 | if attempt < max_retries - 1: 261 | time.sleep(2 ** attempt) # 指数退避 262 | 263 | # 所有重试都失败后返回错误 264 | return Response(f"请求失败,所有重试均未成功。最后的错误: {last_error}", status=500) 265 | 266 | @app.route('/', methods=['GET']) 267 | def home(): 268 | """健康检查端点""" 269 | config_info = { 270 | "status": "Chutes API Service Running", 271 | "version": "1.0", 272 | "has_auth_token": bool(auth_token), 273 | "has_cf_clearance": bool(current_cf_clearance) 274 | } 275 | return config_info 276 | 277 | @app.route('/v1/models', methods=['GET']) 278 | def get_models(): 279 | """获取可用模型列表""" 280 | if not check_auth(): 281 | return Response("Unauthorized", status=401) 282 | 283 | models = [] 284 | for model_id in MODEL_MAPPING.keys(): 285 | models.append({ 286 | "id": model_id, 287 | "object": "model", 288 | "created": int(time.time()), 289 | "owned_by": "chutes" 290 | }) 291 | 292 | return jsonify({ 293 | "object": "list", 294 | "data": models 295 | }) 296 | 297 | @app.route('/v1/chat/completions', methods=['POST']) 298 | def chat(): 299 | """聊天完成接口""" 300 | try: 301 | if not check_auth(): 302 | return Response("Unauthorized", status=401) 303 | 304 | openai_request = request.json 305 | logging.info("收到新的聊天请求") 306 | response = make_request_with_retry(openai_request) 307 | 308 | # 如果返回的是错误响应,直接返回 309 | if isinstance(response, Response): 310 | return response 311 | 312 | # 处理非流式请求 313 | if not openai_request.get('stream', False): 314 | result = process_non_stream_response(response, openai_request.get('model')) 315 | return Response( 316 | json.dumps(result, ensure_ascii=False), 317 | status=200, 318 | content_type='application/json' 319 | ) if isinstance(result, dict) else result 320 | 321 | # 处理流式请求 322 | def generate(): 323 | try: 324 | for line in response.iter_lines(): 325 | if line: 326 | line = line.decode('utf-8') 327 | if line.startswith("data: "): 328 | data = line[6:] 329 | if data == "[DONE]": 330 | yield "data: [DONE]\n\n" 331 | break 332 | 333 | try: 334 | chunk = json.loads(data) 335 | content = process_chunk(chunk) 336 | if content: 337 | response_chunk = { 338 | "id": str(uuid.uuid4()), 339 | "object": "chat.completion.chunk", 340 | "created": int(time.time()), 341 | "model": openai_request.get('model'), 342 | "choices": [{ 343 | "delta": { 344 | "content": content 345 | }, 346 | "index": 0, 347 | "finish_reason": None 348 | }] 349 | } 350 | yield f"data: {json.dumps(response_chunk, ensure_ascii=False)}\n\n" 351 | except json.JSONDecodeError: 352 | continue 353 | 354 | except Exception as e: 355 | logging.error(f"生成响应时出错: {str(e)}", exc_info=True) 356 | return 357 | 358 | return Response( 359 | stream_with_context(generate()), 360 | content_type='text/event-stream' 361 | ) 362 | 363 | except Exception as e: 364 | logging.error(f"聊天接口出错: {str(e)}", exc_info=True) 365 | return Response(f"服务器内部错误: {str(e)}", status=500) 366 | 367 | if __name__ == '__main__': 368 | port = int(os.getenv('PORT', 8805)) 369 | app.run(host='0.0.0.0', port=port, debug=False) 370 | -------------------------------------------------------------------------------- /win/app.py: -------------------------------------------------------------------------------- 1 | from flask import Flask, request, Response, stream_with_context, jsonify 2 | import cloudscraper 3 | import json 4 | import uuid 5 | from datetime import datetime, timezone 6 | import time 7 | import os 8 | 9 | app = Flask(__name__) 10 | 11 | # 模型映射字典 12 | MODEL_MAPPING = { 13 | "nvidia/Llama-3.1-405B-Instruct-FP8": "chutes-nvidia-llama-3-1-405b-instruct-fp8", 14 | "deepseek-ai/DeepSeek-R1": "chutes-deepseek-ai-deepseek-r1", 15 | "Qwen/Qwen2.5-72B-Instruct": "chutes-qwen-qwen2-5-72b-instruct", 16 | "Qwen/Qwen2.5-Coder-32B-Instruc": "chutes-qwen-qwen2-5-coder-32b-instruct", 17 | "bytedance-research/UI-TARS-72B-DPO": "chutes-bytedance-research-ui-tars-72b-dpo", 18 | "OpenGVLab/InternVL2_5-78B": "chutes-opengvlab-internvl2-5-78b", 19 | "hugging-quants/Meta-Llama-3.1-70B-Instruct-AWQ-INT4": "chutes-hugging-quants-meta-llama-3-1-70b-instruct-awq-int4", 20 | "NousResearch/Hermes-3-Llama-3.1-8B": "cxmplexbb-nousresearch-hermes-3-llama-3-1-8b", 21 | "Qwen/QVQ-72B-Preview": "chutes-qwen-qvq-72b-preview", 22 | "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B": "chutes-deepseek-ai-deepseek-r1-distill-qwen-32b", 23 | "jondurbin/bagel-8b-v1.0": "chutes-jondurbin-bagel-8b-v1-0", 24 | "unsloth/QwQ-32B-Preview": "cxmplexbb-unsloth-qwq-32b-preview", 25 | "Qwen/QwQ-32B-Preview": "chutes-qwq-32b-preview", 26 | "jondurbin/airoboros-34b-3.3": "chutes-jondurbin-airoboros-34b-3-3", 27 | "NovaSky-AI/Sky-T1-32B-Preview": "chutes-novasky-ai-sky-t1-32b-preview", 28 | "driaforall/Dria-Agent-a-3B": "chutes-driaforall-dria-agent-a-3b", 29 | "NousResearch/Nous-Hermes-Llama2-13b": "cxmplexbb-nousresearch-nous-hermes-llama2-13b", 30 | "unsloth/Llama-3.2-1B-Instruct": "chutes-unsloth-llama-3-2-1b-instruct" 31 | } 32 | 33 | def check_auth(): 34 | """检查认证""" 35 | auth_token = os.getenv('AUTH_TOKEN') 36 | if not auth_token: 37 | return True 38 | 39 | request_token = request.headers.get('Authorization', '') 40 | return request_token == f"Bearer {auth_token}" 41 | 42 | def create_scraper(): 43 | scraper = cloudscraper.create_scraper( 44 | browser={ 45 | 'browser': 'chrome', 46 | 'platform': 'linux', 47 | 'desktop': True, 48 | 'mobile': False, 49 | 'version': '121.0.0.0' 50 | } 51 | ) 52 | 53 | # 设置请求头 54 | scraper.headers.update({ 55 | "Accept": "*/*", 56 | "Accept-Language": "en-US,en;q=0.9", 57 | "Cache-Control": "no-cache", 58 | "Content-Type": "text/plain;charset=UTF-8", 59 | "Origin": "https://chutes.ai", 60 | "Pragma": "no-cache", 61 | "Referer": "https://chutes.ai/", 62 | "Sec-Ch-Ua": '"Not A(Brand";v="99", "Google Chrome";v="121", "Chromium";v="121"', 63 | "Sec-Ch-Ua-Mobile": "?0", 64 | "Sec-Ch-Ua-Platform": '"Linux"', 65 | "Sec-Fetch-Dest": "empty", 66 | "Sec-Fetch-Mode": "cors", 67 | "Sec-Fetch-Site": "same-origin", 68 | "User-Agent": "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/121.0.0.0 Safari/537.36" 69 | }) 70 | 71 | return scraper 72 | 73 | 74 | def create_chutes_request(openai_request): 75 | """将OpenAI格式请求转换为Chutes格式""" 76 | messages = openai_request['messages'] 77 | message_id = str(uuid.uuid4()) 78 | current_time = datetime.now(timezone.utc).strftime('%Y-%m-%dT%H:%M:%S.%f')[:-3] + 'Z' 79 | 80 | model = openai_request.get('model', 'deepseek-ai/DeepSeek-R1') 81 | chute_name = MODEL_MAPPING.get(model, 'chutes-deepseek-ai-deepseek-r1') 82 | 83 | return { 84 | "messages": [{ 85 | "role": messages[-1]['role'], 86 | "content": messages[-1]['content'], 87 | "id": message_id, 88 | "createdOn": current_time 89 | }], 90 | "model": model, 91 | "chuteName": chute_name 92 | } 93 | 94 | def process_chunk(chunk): 95 | """处理响应数据块""" 96 | try: 97 | if "choices" in chunk and chunk["choices"][0]["delta"].get("content"): 98 | return chunk["choices"][0]["delta"]["content"] 99 | return None 100 | except: 101 | return None 102 | 103 | def process_non_stream_response(response, model): 104 | """处理非流式响应""" 105 | try: 106 | full_content = "" 107 | for line in response.iter_lines(): 108 | if line: 109 | line = line.decode('utf-8') 110 | if line.startswith("data: "): 111 | data = line[6:] 112 | if data == "[DONE]": 113 | break 114 | try: 115 | chunk = json.loads(data) 116 | content = process_chunk(chunk) 117 | if content: 118 | full_content += content 119 | except json.JSONDecodeError: 120 | continue 121 | 122 | if not full_content: 123 | return Response("Empty response from server", status=500) 124 | 125 | return { 126 | "id": str(uuid.uuid4()), 127 | "object": "chat.completion", 128 | "created": int(time.time()), 129 | "model": model, 130 | "choices": [{ 131 | "message": { 132 | "role": "assistant", 133 | "content": full_content 134 | }, 135 | "finish_reason": "stop", 136 | "index": 0 137 | }], 138 | "usage": { 139 | "prompt_tokens": 0, 140 | "completion_tokens": 0, 141 | "total_tokens": 0 142 | } 143 | } 144 | except Exception as e: 145 | print(f"Error processing non-stream response: {str(e)}") 146 | return Response("Failed to process response", status=500) 147 | 148 | @app.route('/', methods=['GET']) 149 | def home(): 150 | """健康检查端点""" 151 | return {"status": "Chutes API Service Running", "version": "1.0"} 152 | 153 | @app.route('/v1/models', methods=['GET']) 154 | def get_models(): 155 | """获取可用模型列表""" 156 | if not check_auth(): 157 | return Response("Unauthorized", status=401) 158 | 159 | models = [] 160 | for model_id in MODEL_MAPPING.keys(): 161 | models.append({ 162 | "id": model_id, 163 | "object": "model", 164 | "created": int(time.time()), 165 | "owned_by": "chutes" 166 | }) 167 | 168 | return jsonify({ 169 | "object": "list", 170 | "data": models 171 | }) 172 | 173 | @app.route('/v1/chat/completions', methods=['POST']) 174 | def chat(): 175 | """聊天完成接口""" 176 | try: 177 | if not check_auth(): 178 | return Response("Unauthorized", status=401) 179 | 180 | openai_request = request.json 181 | chutes_request = create_chutes_request(openai_request) 182 | scraper = create_scraper() 183 | 184 | headers = { 185 | "Accept": "*/*", 186 | "Accept-Language": "zh-CN,zh;q=0.9,en;q=0.8", 187 | "Cache-Control": "no-cache", 188 | "Content-Type": "text/plain;charset=UTF-8", 189 | "Origin": "https://chutes.ai", 190 | "Pragma": "no-cache", 191 | "Referer": "https://chutes.ai/app/chute/590d919c-8d4c-5b7b-9445-ed2cd71944a8", 192 | "Sec-Ch-Ua": '"Not A(Brand";v="8", "Chromium";v="132", "Microsoft Edge";v="132"', 193 | "Sec-Ch-Ua-Mobile": "?0", 194 | "Sec-Ch-Ua-Platform": '"Windows"', 195 | "Sec-Fetch-Dest": "empty", 196 | "Sec-Fetch-Mode": "cors", 197 | "Sec-Fetch-Site": "same-origin" 198 | } 199 | 200 | response = scraper.post( 201 | "https://chutes.ai/app/api/chat", 202 | headers=headers, 203 | json=chutes_request, 204 | stream=True 205 | ) 206 | 207 | if response.status_code != 200: 208 | return Response(f"Chutes API error: {response.text}", status=response.status_code) 209 | 210 | # 处理非流式请求 211 | if not openai_request.get('stream', False): 212 | result = process_non_stream_response(response, chutes_request["model"]) 213 | return Response( 214 | json.dumps(result, ensure_ascii=False), 215 | status=200, 216 | content_type='application/json' 217 | ) if isinstance(result, dict) else result 218 | 219 | # 处理流式请求 220 | def generate(): 221 | try: 222 | for line in response.iter_lines(): 223 | if line: 224 | line = line.decode('utf-8') 225 | if line.startswith("data: "): 226 | data = line[6:] 227 | if data == "[DONE]": 228 | yield "data: [DONE]\n\n" 229 | break 230 | 231 | try: 232 | chunk = json.loads(data) 233 | content = process_chunk(chunk) 234 | if content: 235 | response_chunk = { 236 | "id": str(uuid.uuid4()), 237 | "object": "chat.completion.chunk", 238 | "created": int(time.time()), 239 | "model": chutes_request["model"], 240 | "choices": [{ 241 | "delta": { 242 | "content": content 243 | }, 244 | "index": 0, 245 | "finish_reason": None 246 | }] 247 | } 248 | yield f"data: {json.dumps(response_chunk, ensure_ascii=False)}\n\n" 249 | except json.JSONDecodeError: 250 | continue 251 | 252 | except Exception as e: 253 | print(f"Error in generate: {str(e)}") 254 | return 255 | 256 | return Response( 257 | stream_with_context(generate()), 258 | content_type='text/event-stream' 259 | ) 260 | 261 | except Exception as e: 262 | print(f"Error in chat endpoint: {str(e)}") 263 | return Response(f"Internal server error: {str(e)}", status=500) 264 | 265 | if __name__ == '__main__': 266 | port = int(os.getenv('PORT', 8805)) 267 | app.run(host='0.0.0.0', port=port, debug=False) 268 | --------------------------------------------------------------------------------