添加模型的时候能够选择是视觉模型还是文本模型

[Xheldon]

• 我已确认目前没有类似 issue
• 我已确认我已升级到最新版本
• 我已完整浏览项目 README 和项目文档并未找到解决方案
• 我理解并愿意跟进此 issue，协助测试和提供反馈
• 我将以礼貌和尊重的态度提问，不得使用不文明用语 (包括在此发布评论的所有人同样适用, 不遵守的人将被 block)
• 我理解并认可上述内容，并理解项目维护者精力有限，不遵循规则的 issue 可能会被无视或直接关闭

功能描述

现在有挺多视觉模型，比如 Qwen/Qwen2-VL-72B-Instruct 和 grok-vision-beta，我添加模型并使用后，发送的消息依然是 message 为字符串，并未识别为视觉模型，是我用法不对还是说某个模型的类型是文本还是视觉，需要大佬挨个适配？

feature 需求：可否在添加模型的时候选择当前模型是否为视觉模型还是文本？

相关截图 (如果有)

[Issues-translate-bot]

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

---

Title: When adding a model, you can choose whether it is a visual model or a text model

• I have confirmed that there is currently no similar issue
• I have confirmed that I have upgraded to the latest version
• I have fully browsed the project README and project documentation and did not find a solution
• I understand and am willing to follow up on this issue, assist in testing and provide feedback
• I will ask questions in a polite and respectful manner and shall not use uncivilized language (the same applies to everyone who posts comments here, those who do not comply will be blocked)
• I understand and acknowledge the above content, and understand that project maintainers have limited energy. Issues that do not follow the rules may be ignored or closed directly

Function description

There are many visual models now, such as Qwen/Qwen2-VL-72B-Instruct and grok-vision-beta. After I added the model and used it, the message sent was still a string and was not recognized as a visual model. It was me. Is the usage wrong? Or is it that the type of a certain model is text or visual, and the boss needs to adapt it one by one?

Feature requirement: Can I choose whether the current model is a visual model or text when adding a model?

Relevant screenshots (if any)

[Xheldon]

我病急乱投医把映射了一下也还是不行。。模型映射是这样写的：!gpt-4-vision-preview>grok-vision-beta
我理解模型映射的作用是调用的形式跟服务端实际请求的存在差异所以需要这个东西，所以我前端调 gpt-4-vision-preview 实际服务端请求 grok-vision-beta 但还是不行 !gpt-4-1106-vision-preview>grok-vision-beta 同样。

是模型不支持 url 图片读取还是怎么回事呢

[Issues-translate-bot]

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

---

I was so sick that I went to the doctor and tried mapping, but it still didn't work. . The model mapping is written like this:!gpt-4-vision-preview>grok-vision-beta
I understand that the role of model mapping is that there is a difference between the calling form and the actual request from the server, so this thing is needed, so I adjusted gpt-4-vision-preview on the front end and the actual server requested grok-vision-beta, but it still didn’t work! gtp-4 -vision-preview>grok-vision-beta Same.

Is it because the model does not support url image reading or something else is going on?

[zmh-program]

这样映射作为解决方案是可以的，如果工作不正常请开启后台调试模式，并提供请求体。

在长期解决来看，企业版提供内置模型表并且可自定义模型类型，后续会同步此架构设计。

[Issues-translate-bot]

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

---

This mapping can be used as a solution. If it does not work properly, please enable background debugging mode and provide the request body.
The open source version does not design custom model types and will be synchronized later.

[Xheldon]

之前因为 watchtower 的的原因（我提过的另一个 issue），所以使用了稳定版 stable 的 docker，它似乎没有 调试模式。。。
我等你们同步架构设计吧，先不整这个了

[Issues-translate-bot]

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

---

Previously, because of watchtower (another issue I mentioned), I used the stable version of docker, which seems to have no debug mode. . .
I'll wait for you to synchronize the architecture design, let's not deal with this yet.