ChatGPTNextWeb
[Bug] Authentication Fails in NextChat + One-API Local Deployment When Using DeepSeek Official Model & API
📦 Deployment Method
Docker
📌 Version
2.6.0
💻 Operating System
Windows
📌 System Version
win10
🌐 Browser
Chrome
📌 Browser Version
Latest
🐛 Bug Description
When deploying NextChat with One-API locally, an "Authentication Fails (no such user)" error occurs when attempting to use the DeepSeek official model. The issue arises because NextChat does not correctly route requests through One-API's gateway, instead attempting to directly call DeepSeek's API without using the One-API Token. Other UI applications (e.g., Chatbox) work correctly with the same setup.
Error Message:
JSON
{
"error": {
"message": "Authentication Fails (no such user)",
"type": "authentication_error",
"param": null,
"code": "invalid_request_error"
}
}
📷 Recurrence Steps
- Deploy NextChat and One-API locally.
- Add DeepSeek's API key and other channels (e.g., Alibaba's Tongyi Qianwen) in One-API.
- Configure NextChat's environment variables with the One-API token as follows: environment:
- OPENAI_API_KEY=sk-dUJhlhHAabdVXsqN22B205Bxxxxxxxxxxxxx6e65B1Ad
- CODE=xxxxx
- BASE_URL=http://154.xxx.241.166:3000
- CUSTOM_MODELS=-all,+deepseek-ai/DeepSeek-R1,+Pro/deepseek-ai/DeepSeek-R1,+deepseek-ai/DeepSeek-V3,+Pro/deepseek-ai/DeepSeek-V3,+deepseek-r1,+deepseek-v3,+qwen-max-latest,+deepseek-r1-250120,+deepseek-v3-241226,deepseek-chat,deepseek-reasoner
- DEFAULT_MODEL=deepseek-chat
🚦 Expected Behavior
NextChat should route requests through One-API's gateway and successfully connect to the DeepSeek model using the One-API Token.
Actual Behavior: An "Authentication Fails (no such user)" error occurs. The logs indicate that NextChat is attempting to directly call DeepSeek's API without using the One-API Token.
📝 Additional Information
The issue is similar to a previously reported bug where NextChat failed to route requests correctly for the SiliconFlow model. Other UI applications (e.g., Chatbox) work correctly with the same One-API setup, indicating that the issue is specific to NextChat's implementation.
Proposed Solution:
Ensure that NextChat correctly routes requests through One-API's gateway and uses the One-API Token for authentication. Verify that the BASE_URL and OPENAI_API_KEY environment variables are correctly configured in NextChat. Update NextChat's request handling logic to properly utilize One-API's gateway for all model requests.
同样问题👀也是部署 one-api，看了日志发现调用 DeepSeek 官方的 API 地址去了。
NextChat 聊天报错如下：
按如下更换名称，报错又变了。
CUSTOM_MODELS="-all,+deepseek-chat@DeepSeek💰=DeepSeek-V3,+deepseek-reasoner@DeepSeek💰=DeepSeek-R1"
同样的报错，通过one-api转发官方模型，以前是正常的，docker更新至v2.16.0（实际显示v2.15.8）后就不正常了。 后来在one-api中设置模型重定向就通了，CUSTOM_MODELS中不要指定@provider。
Bot detected the issue body's language is not English, translate it automatically.
The same error is reported, and the official model forwarding through one-api is normal before, but it is abnormal after docker is updated to v2.16.0 (actually displayed v2.15.8). Later, the model redirection is set in one-api and it will be enabled. Do not specify @provider in CUSTOM_MODELS.
