[Bug] 404 Error when adding Embedding Model

[rainy0710]

Contact Information

xyzhang_cn@outlook.com

MaxKB Version

1.10.2-lts

Problem Description

Steps to Reproduce

404 error occurs when Adding Embedding Model "bge-m3:567m" to MaxKB by Ollama.

The expected correct result

Successfully adding the "bge-m3:567m" model to MaxKB by Ollama.

Related log output

Additional Information

No response

[Shenguobin0102]

你好，看起来是你的url有问题，我们的环境添加这个模型是正常的，请检查你的 ollama 环境，看看其他模型是否能正常添加。

[shaohuzhang1]

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

---

Hello, it seems that there is a problem with your url. It is normal to add this model in our environment. Please check your ollama environment to see if other models can be added normally.

[rainy0710]

你好，看起来是你的url有问题，我们的环境添加这个模型是正常的，请检查你的 ollama 环境，看看其他模型是否能正常添加。

你好，感谢答疑！

不过我使用部署在同一台服务器上的dify来加载该模型时，显示没有问题，URL和模型名称都一致，但是MaxKB这边汇报404：

[shaohuzhang1]

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

---

Hello, it seems that there is a problem with your url. It is normal to add this model in our environment. Please check your ollama environment to see if other models can be added normally.

Hello, thanks for the Q&A!

However, when I use diify deployed on the same server to load the model, there is no problem, the URL and model name are the same, but MaxKB reports 404:

[rainy0710]

你好，看起来是你的url有问题，我们的环境添加这个模型是正常的，请检查你的 ollama 环境，看看其他模型是否能正常添加。

重新尝试了一下，其他大语言模型都没有问题，就是向量模型会报错，换了一个向量模型一样报错：

然后我尝试将本地ollama模型删掉，然后再在maxkb中加载，它可以成功执行模型下载操作，最后成功添加：

但是我重新添加上一步刚下载好的这个向量模型时，还是一样的错误：

也就是说，可以添加本地不存在的向量模型（自动调用Ollama pull，然后成功添加模型），但是直接添加本地存在的向量模型就会报错，我查看了一下报错请求日志，错误信息如下：

[shaohuzhang1]

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

---

Hello, it seems that there is a problem with your url. It is normal to add this model in our environment. Please check your ollama environment to see if other models can be added normally.

After trying again, there is no problem with other large language models, just the vector model will report an error, and if you change to a vector model, you will report an error:

Then I tried to delete the local ollama model and then load it in maxkb, which can successfully perform the model download operation and finally successfully add it:

But when I added the vector model I just downloaded in the previous step, the same error was still:

In other words, you can add a vector model that does not exist locally (automatically call Ollama pull, and then successfully add the model), but adding a locally existing vector model will cause an error. I checked the error request log, and the error message is as follows:

[Shenguobin0102]

你好，我们按照你描述的场景做了复现，但是没有复现你描述的问题，请你再确认一下你环境中模型的名称是不是添加时填入的名称。

[shaohuzhang1]

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

---

Hello, we reproduced the scenario you described, but did not reproduce the problem you described. Please confirm whether the name of the model in your environment is the name you filled in when adding it.