model-mapper和transformer插件兼容问题 || Compatibility issues with model-mapper and transformer plug-in

[yumc2573]

同时开启model-mapper和transformer插件，会导致model-mapper插件功能失效，直接使用原始请求body的model参数请求大模型，如下：

1.请求body：
{
"model": "deepseek-r1-250120",
"stream": false,
"messages": [
{
"role": "user",
"content": "10个字介绍下淮安"
}
]
}

2.路由配置：

3.transformer插件配置：

4.model-mapper配置：

5.请求日志：

预期结果是：会使用“deepseek-r11”去请求大模型。

---

Turn on the model-mapper and transformer plug-ins at the same time will cause the model-mapper plug-in to fail. You can directly use the model parameters of the original request body to request the big model, as follows:

1. Request body:
   {
   "model": "deepseek-r1-250120",
   "stream": false,
   "messages": [
   {
   "role": "user",
   "content": "Introduced by Huai'an in 10 words"
   }
   ]
   }

2.Routing configuration:

3. Transformer plug-in configuration:

4. Model-mapper configuration:

5. Request log:

The expected result is: "deepseek-r11" will be used to request the big model.

[CH3CHO]

cc @johnlanni @WeixinX @Uncle-Justice @rinfx

[CH3CHO]

这个也请 @WeixinX 看一下

---

Please check this @WeixinX for a look

[WeixinX]

@yumc2573 hello，我粗浅理解你的意思是，从 body 里面获取 model name (r11)，赋值到 x-model (r11)，以匹配对应请求头（r1-250120），显然两者冲突了，因此想问下你实际的意图是什么？

---

@yumc2573 hello, I understand what you mean is to get the model name (r11) from the body and assign it to x-model (r11) to match the corresponding request header (r1-250120). Obviously the two conflict, so I would like to ask what your actual intention is?

[WeixinX]

@yumc2573 另外，model-mapper 插件的执行阶段为 AUTHN，优先级为 800，而 transform 插件的执行阶段同为 AUTHN，优先级是 410。换句话说，model-mapper 会先于 transform 执行，是符合预期的。若有需要，可尝试自行修改两者的 WasmPlugin 资源。

---

@yumc2573 In addition, the execution stage of the model-mapper plug-in is AUTHN with a priority of 800, while the execution stage of the transform plug-in is AUTHN with a priority of 410. In other words, model-mapper will execute before transform, which is as expected. If necessary, try modifying the WasmPlugin resources of both yourself.

[yumc2573]

@WeixinX 你好，我的场景是，比如模型deepseek-r1的AI路由中配置不同供应商的资源，豆包的（官方model id是deepseek-r1-xxx）,azure的（官方model id是DeepSeek-R1），然后我暴露给我用户的model id是DeepSeek-R1，也就是用户可以通过DeepSeek-R1来使用我配置的deepseek-r1的服务，但是结果是请求走到azure是正常的，但是走的豆包时并不是用的deepseek-r1-xxx去请求的，而是用的请求body的DeepSeek-R1去请求的。

后来我放弃了transform插件，改成model-router插件就ok了，而上面的问题我理解应该是插件之间有什么冲突吧。

---

@WeixinX Hello, my scenario is, for example, the AI ​​route of model deepseek-r1 is configured with resources from different vendors, bean bun (the official model id is deepseek-r1-xxx), and azure (the official model id is DeepSeek-R1), and then I exposed the model id to my user is DeepSeek-R1, which means that the user can use the deepseek-r1 service I configured through DeepSeek-R1, but the result is that it is normal to request it to Azure, but when I go to the bean bun, I do not use deepseek-r1-xxx to request it, but the DeepSeek-R1 request it.

Later I gave up the transform plug-in and changed it to the model-router plug-in and it was OK. I understand the above problem that should be the conflict between the plug-ins.

[WeixinX]

理解了，它们没有冲突，只是执行有先后顺序：model-router > model-mapper > transform

---

Understand that they have no conflicts, they just have order of execution: model-router > model-mapper > transform