[Bug] 第三方 gpt 代理无法使用 o1\o1mini

[LarkinZero]

📦 部署环境

Docker

📌 软件版本

1.17.2

💻 系统环境

Windows, macOS

🌐 浏览器

Edge

🐛 问题描述

日志

Route: [openai] InvalidProviderAPIKey:  [Error: 401 status code (no body)] {
  status: 401,
  headers: {
  access-control-allow-origin: '*',
  connection: 'keep-alive',
  content-length: '281',
  content-type: 'application/json; charset=utf-8',
  date: 'Sat, 14 Sep 2024 01:34:33 GMT',
  server: 'nginx/1.18.0 (Ubuntu)',
  vary: 'Origin'
},
  request_id: undefined,
  error: undefined,
  code: undefined,
  param: undefined,
  type: undefined
}

📷 复现步骤

没有修改模型api key 的情况下，通过刷新模型列表，获取到支持了 o1 然后切换到o1 模型后，发起对话，报错提示OpenAI API Key 不正确或为空，请检查 OpenAI API Key 后重试 重试依旧报错

🚦 期望结果

No response

📝 补充信息

No response

[lobehubbot]

👀 @LarkinZero

Thank you for raising an issue. We will investigate into the matter and get back to you as soon as possible. Please make sure you have given us as much context as possible.\ 非常感谢您提交 issue。我们会尽快调查此事，并尽快回复您。 请确保您已经提供了尽可能多的背景信息。

[dosubot[bot]]

@LarkinZero 你好！我是Dosu，我是一个可以帮助您解决错误，回答问题，并成为贡献者的机器人。我会在等待人类维护者时帮助您。

I found some similar issues and discussions that might help resolve your problem:

1. [Bug] 我使用gpt-4o-mini对话提示openai密钥错误: This issue was resolved by noting that the interface proxy did not support the new model ^[1].
2. [Bug] 新会话总是提示“OpenAI API Key 不正确或为空，请检查 OpenAI API Key 后重试”，即便默认的模型不是chatgpt: The solution was to add provider=ollama ^[2].
3. [Bug] OpenAI API Key 为空，请添加自定义 OpenAI API Key: This issue was closed without a specific solution provided ^[3].

For your specific issue with the o1 and o1mini models, you need to ensure that the payload for these models is pruned using the pruneO1Payload function. This function sets frequency_penalty and presence_penalty to 0, changes the role of messages from 'system' to 'user', disables streaming, and sets temperature and top_p to 1. Here is the relevant code snippet:

export const o1Models = new Set([
  'o1-preview',
  'o1-preview-2024-09-12',
  'o1-mini',
  'o1-mini-2024-09-12',
]);

export const pruneO1Payload = (payload: ChatStreamPayload) => ({
  ...payload,
  frequency_penalty: 0,
  messages: payload.messages.map((message: OpenAIChatMessage) => ({
    ...message,
    role: message.role === 'system' ? 'user' : message.role,
  })),
  presence_penalty: 0,
  stream: false,
  temperature: 1,
  top_p: 1,
});

export const LobeOpenAI = LobeOpenAICompatibleFactory({
  baseURL: 'https://api.openai.com/v1',
  chatCompletion: {
    handlePayload: (payload) => {
      const { model } = payload;

      if (o1Models.has(model)) {
        return pruneO1Payload(payload) as any;
      }

      return { ...payload, stream: payload.stream ?? true };
    },
  },
  debug: {
    chatCompletion: () => process.env.DEBUG_OPENAI_CHAT_COMPLETION === '1',
  },
  provider: ModelProvider.OpenAI,
});

Make sure your implementation includes this function to handle the payload correctly for the o1 and o1mini models ^[4].

Additionally, ensure that you have correctly set the OPENAI_API_KEY environment variable with a valid API key. Here is the Docker command to run LobeChat with the correct environment variables:

$ docker run -d -p 3210:3210 \
  -e OPENAI_API_KEY=sk-xxxx \
  -e OPENAI_PROXY_URL=https://api-proxy.com/v1 \
  -e ACCESS_CODE=lobe66 \
  --name lobe-chat \
  lobehub/lobe-chat

Replace sk-xxxx with your actual OpenAI API key. If you are using a third-party proxy, ensure that the OPENAI_PROXY_URL is correctly set to the proxy address provided by the third-party service ^[5].

To continue talking to Dosu, mention @dosu.

[LarkinZero]

已解决，是 api key 的权限没有更新。需要手动在 api 提供商的后台将权限刷新即可

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Solved, the permissions of the api key are not updated. You need to manually refresh the permissions in the api provider's backend

[lobehubbot]

✅ @LarkinZero

This issue is closed, If you have any questions, you can comment and reply.\ 此问题已经关闭。如果您有任何问题，可以留言并回复。