[Bug]:v0.17.1 Add xinference model error Connection error

[ran411285752]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (Language Policy).
• Non-english title submitions will be closed directly ( 非英文标题的提交将会被直接关闭 ) (Language Policy).
• Please do not modify this template :) and fill in all the required fields.

RAGFlow workspace code commit ID

d447392

RAGFlow image version

d447392(v0.17.1)

Other environment information

Actual behavior

使用v0.17.0，添加xinference运行的本地模型，是没有问题的
但是使用v0.17.1，添加xinference运行的本地模型，同样的操作和配置，会报错Connection error

Expected behavior

使用v0.17.1，可以正常添加xinference运行的本地模型

Steps to reproduce

1、使用v0.17.1的docker镜像运行ragflow
2、添加llm模型，使用xinference供应商
3、只要是v0.17.1添加，无论是什么类型的模型，都会报错Connection error.

Additional information

报错日志：
2025-03-12 17:41:56,913 INFO 19 Retrying request to /chat/completions in 0.782334 seconds
2025-03-12 17:41:57,714 INFO 19 Retrying request to /chat/completions in 1.503957 seconds
2025-03-12 17:41:59,230 ERROR 19
Fail to access model(qwen2.5-instruct).ERROR: Connection error.
NoneType: None

2025-03-12 17:58:43,702 INFO 19 Retrying request to /embeddings in 0.873221 seconds
2025-03-12 17:58:44,579 INFO 19 Retrying request to /embeddings in 1.891999 seconds
2025-03-12 17:58:46,477 ERROR 19
Fail to access embedding model(Conan-embedding-v1).Connection error.
NoneType: None

[LongMarching]

Same issue. v0.17.1-slim unable to add embedding models deployed by xinference. Rerank model which also deployed by xinference is correctly added(can't select in System Model Settings, but it can be called using the curl command in PowerShell.)

[Simonqujian78]

yes，I have the same question.all models add through ollama would not be shown in the list of system setting,include chat,embedding. Ollama shows in model provider. click "more" also shows the models,but they are not in the list for selection.

[KevinHuSh]

What about filling the API-Key with something while adding models?

[zyddsg159357]

What about filling the API-Key with something while adding models?

this is useful, thank you

[ran411285752]

v0.17.2已解决该bug