[Feature Request] 利用自定义接口和模型方式使用阿里DeepSeek-R1模型时无思考流程

[Alauda95]

🥰 需求描述

使用阿里云API
base url为https://dashscope.aliyuncs.com/compatible-mode/v1
mode名称为
deepseek-r1
deepseek-v3

🧐 解决方案

增加适配

📝 补充信息

No response

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

Title: [Feature Request] Alibaba provides API calls to the DeepSeek model

🥰 Requirement description

Using Alibaba Cloud API
The base url is https://dashscope.aliyuncs.com/compatible-mode/v1
The mode name is
deepseek-r1
deepseek-v3

🧐 Solution

Added adaptation

📝 Supplementary information

No response

[yangxiang92]

把base url里面的v1去掉才可以。
也就是设置成：https://dashscope.aliyuncs.com/compatible-mode

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

Remove the v1 in the base url.
That is to set to: https://dashscope.aliyuncs.com/compatible-mode

[wangfan002]

但是UI不能显示思考过程吧

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

But the UI can't show the thinking process

[Alauda95]

把base url里面的v1去掉才可以。 也就是设置成：https://dashscope.aliyuncs.com/compatible-mode

自定义接口，前端发送。
添加配置，后端发送。
都可，只是个建议，前者麻烦一点点。

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

Remove the v1 in the base url. That is to set to: https://dashscope.aliyuncs.com/compatible-mode

Custom interface, front-end send.
Add configuration, send backend.
All is OK, just a suggestion, the former is a little troublesome.

[Alauda95]

但是UI不能显示思考过程吧

我这边是的。自定义模型使用阿里deepseek-r1，控制台（F12）显示请求在接受思考过程，但UI显示等待。

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

But the UI can't show the thinking process

I'm here. Custom model uses Alibaba deepseek-r1, console (F12) to display requests during the acceptance thinking process, but the UI shows waiting.

[yangxiang92]

对阿里云百炼平台的DeepSeek的API显示思维链在main分支已经支持了，但是由于很久没有发布版本了，导致当前的Docker版本（2.15.8）是不支持的。

在main分支，通过配置

DEEPSEEK_URL="https://dashscope.aliyuncs.com/compatible-mode/v1"
DEEPSEEK_API_KEY="sk-xxxxx"
CUSTOM_MODELS="+deepseek-r1@DeepSeek"

是可以正常显示r1模型的思维链的。

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

The DeepSeek API for Alibaba Cloud Bailian platform shows that the thinking chain has been supported in the main branch, but since the version has not been released for a long time, the current Docker version (2.15.8) is not supported.

On the main branch, configure

DEEPSEEK_URL="https://dashscope.aliyuncs.com/compatible-mode/v1"
DEEPSEEK_API_KEY="sk-xxxxx"
CUSTOM_MODELS="+deepseek-r1@DeepSeek"

It can display the thinking chain of the r1 model normally.

[csliuchi]

对阿里云百炼平台的DeepSeek的API显示思维链在main分支已经支持了，但是由于很久没有发布版本了，导致当前的Docker版本（2.15.8）是不支持的。

在main分支，通过配置

DEEPSEEK_URL="https://dashscope.aliyuncs.com/compatible-mode/v1" DEEPSEEK_API_KEY="sk-xxxxx" CUSTOM_MODELS="+deepseek-r1@DeepSeek"

是可以正常显示r1模型的思维链的。

请问字节火山引擎的也可以通过类似配置来支持吗？

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

The DeepSeek API for Alibaba Cloud Bailian platform shows that the thinking chain has been supported in the main branch, but since the version has not been released for a long time, the current Docker version (2.15.8) is not supported.

In main branch, configure

DEEPSEEK_URL="https://dashscope.aliyuncs.com/compatible-mode/v1" DEEPSEEK_API_KEY="sk-xxxxx" CUSTOM_MODELS="+deepseek-r1@DeepSeek"

It can display the thinking chain of the r1 model normally.

Can the Byte Volcano Engine be supported through similar configurations?

[yangxiang92]

对阿里云百炼平台的DeepSeek的API显示思维链在main分支已经支持了，但是由于很久没有发布版本了，导致当前的Docker版本（2.15.8）是不支持的。
在main分支，通过配置
DEEPSEEK_URL="https://dashscope.aliyuncs.com/compatible-mode/v1" DEEPSEEK_API_KEY="sk-xxxxx" CUSTOM_MODELS="+deepseek-r1@DeepSeek"
是可以正常显示r1模型的思维链的。

请问字节火山引擎的也可以通过类似配置来支持吗？

我猜应该是可以的，查了下提交历史，似乎是只要模型结尾有“r1/R1”字样就会显示思维链。

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

The DeepSeek API for Alibaba Cloud Bailian platform shows that the thinking chain has been supported in the main branch, but since the version has not been released for a long time, the current Docker version (2.15.8) is not supported.
In main branch, configure
DEEPSEEK_URL="https://dashscope.aliyuncs.com/compatible-mode/v1" DEEPSEEK_API_KEY="sk-xxxxx" CUSTOM_MODELS="+deepseek-r1@DeepSeek"
It can display the thinking chain of the r1 model normally.

Can the Byte Volcano Engine be supported through similar configurations?

I guess it should be OK. After checking the submission history, it seems that as long as there is the word "r1/R1" at the end of the model, the thinking chain will be displayed.