由于deepseek-r1的官方API并未完全遵循openai的接口开放规范，导致open-webui无法显示deepseek-r1的思考过程，现在撸了个python脚本，支持了阿里云、百度云（其他未实测）的deepseek-r1在open-webui上的使用，并且可以正常显示思维链
下面是为什么deepseek-r1无法在open-webui上显示思维过程的原因
解决方案
直接上python源码，注意修改备注内容
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 |
import time import uuid import tornado.ioloop import tornado.web import tornado.gen import httpx import json import traceback api_key = 'sk-xxx' # API KEY, 必填, 自己修改 appid = '' # APPID, 百度云可选Header chat_completions_url = 'https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions' class ModelsHandler(tornado.web.RequestHandler): # 用于在open-webui上添加api url过后显示模型列表 async def get(self): self.write({ 'data': [ { "id": "deepseek-r1", "object": "model", "created": 1626777600, "owned_by": "openai", "permission": [{ "id": "modelperm-LwHkVFn8AcMItP432fKKDIKJ", "object": "model_permission", "created": 1626777600, "allow_create_engine": True, "allow_sampling": True, "allow_logprobs": True, "allow_search_indices": False, "allow_view": True, "allow_fine_tuning": False, "organization": "*", "group": None, "is_blocking": False }], "root": "deepseek-r1", "parent": None } ] }) class ChatHandler(tornado.web.RequestHandler): async def get(self): await self.post() async def post(self): self.set_header("Cache-Control", "no-cache") self.set_header("Content-Type", "text/event-stream") try: body_json = json.loads(self.request.body) await self.process_chat(body_json) except Exception as e: print(f"Error in post: {e}") self.set_status(500) self.write({"error": str(e)}) async def process_chat(self, body_json): try: headers = { "Authorization": f"Bearer %s" % api_key, "appid": appid, } async with httpx.AsyncClient(timeout=None) as client: async with client.stream( 'POST', chat_completions_url, headers=headers, json={ "model": "deepseek-r1", "messages": body_json['messages'], "stream": True } ) as response: think_start = False think_end = False write_data = '' chat_id = str(uuid.uuid4()) async for line in response.aiter_lines(): if line.startswith('data: '): try: line = line[6:] if line == '[DONE]': break chunk = json.loads(line) if not chunk: continue delta = chunk['choices'][0]['delta'] reasoning_content = delta.get('reasoning_content', '') or delta.get('model_extra', {}).get('reasoning_content', '') content = delta.get('content', '') if reasoning_content: if not think_start: write_data += '<think>' think_start = True write_data += reasoning_content elif content: if not think_end and think_start: write_data += '</think>' think_end = True write_data += content json_data = { "id": "chatcmpl-%s" % chat_id, "choices": [{ "delta": { "content": write_data, "role": "assistant" }, "finish_reason": None, "index": 0 }], "created": int(time.time()), "model": "deepseek-r1", "object": "chat.completion.chunk", "system_fingerprint": "fp_1dae2613a92e", "usage": None, "provider": "xiaoC" } self.write(f'data: {json.dumps(json_data)}\n\n') await self.flush() write_data = '' except Exception as e: traceback.print_exc() print(f"Error processing chunk: {e}") continue except Exception as e: traceback.print_exc() print(f"Error in process_chat: {e}") raise def make_app(): return tornado.web.Application([ (r"/v1/chat/completions", ChatHandler), (r"/v1/models", ModelsHandler), ]) if __name__ == "__main__": app = make_app() app.listen(8701) print("Listening on http://localhost:8701") tornado.ioloop.IOLoop.current().start() |
在部署open-webui上的机器上把这个python脚本跑起来（可能会提示需要一些包，自己使用python3 -m pip install xxx进行安装，这里不做赘述）
进入open-webui → 管理员面板 → 设置 → 外部链接 → 在管理OpenAI API连接右边点击加号进行添加，输入http://127.0.0.1:8701/v1(注意，如果你是docker容器内部署的open-webui, 这里需要填写docker宿主机的ip，类似于172.17.0.1, 具体地址请使用ip addr查看docker0), 由于密匙已经写死在了python脚本中, 所以这里的密匙请随便填一个，当然你可以将python代码修改为这里传入的秘钥。
确定后，如果没有问题的话，在模型列表就会出现这个模型
选择该模型，确认能正常使用。恩，AI依然还是不会讲笑话。
