Local embedding service (provided by Xinference) already configured as default, though when indexing docs it still connects to OpenAI

[xraywu]

Self Checks

• [X] I have searched for existing issues search for existing issues, including closed ones.
• [X] I confirm that I am using English to file this report (我已阅读并同意 Language Policy).

Dify version

0.3.34

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

1. Set up an Xinference embedding model as default embedding service
2. Upload a document for indexing
3. The indexing process times out and an error of '错误 HTTPSConnection Pool(host='openaipublic.blob.core.windows.net', port=443): Max retriesexceeded with url: /gpt-2/encodings/main/vocab.bpe (Caused by ConnectTimeoutError(<urllib3.connection.HTTPSConnection object at 0x7f84e4c34730>, 'Connection to openaipublic.blob.core.windows.net timed out. (connect timeout=None)')) ' is displayed

✔️ Expected Behavior

It connects to local embedding service and the document can be indexed properly

❌ Actual Behavior

It connects to OpenAI endpoint and caused a timeout

[xraywu]

btw. I've checked this pr e55dd13 but adding environmental variable does not seem to help

[sisuad]

I had the same promble. cause you chatbot already chose openai . You can change openai in you chatbox settings

Seems v0.4.1 already fixed this problem but I can't ensure it cause I have another error

[xraywu]

I had the same promble. cause you chatbot already chose openai . You can change openai in you chatbox settings

Seems v0.4.1 already fixed this problem but I can't ensure it cause I have another error

Not the same issue. Check #1876 , this issue should be resolved in next release

[takatost]

fixed in version 0.4.2