After upgrading, the local deployment model cannot be used in the intranet environment

[opsvr202]

Self Checks

• This is only for bug report, if you would like to ask a question, please head to Discussions.
• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）
• Please do not modify this template :) and fill in all the required fields.

Dify version

0.15.0

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

Environment description: privatization deployment of dify, v0.15.0 version, dify docking with xinference's inference embedding and rerank model
Question 1: The knowledge base embedding cannot use local embedding and rerank models. The screenshot of the problem is as follows

Question 2: The question-and-answer feature cannot utilize the local model: The screenshot of the issue is provided below:

✔️ Expected Behavior

Call the xinference model for question answering and embedding

❌ Actual Behavior

Unable to call xinference model for question answering and embedding

[dosubot]

I found some similar discussions that might be relevant to your issue:

1. Error: Response output is missing or does not contain embeddings - This discussion is unsolved and might provide insights into similar embedding issues ^[1].

2. Xinference failed to access the local model - This is another unsolved discussion that deals with accessing local models, which might be related to your problem ^[2].

For your specific issue, the DNS resolution errors you're encountering suggest that the Dify deployment might be having trouble resolving external domains. Here are some common causes and solutions for connection errors when using local models in Dify:

• Cross-Domain Issues: Ensure that the CORS configuration in docker-compose.yml allows the necessary domains.
• Incorrect Server Address: Use the host's local network IP address instead of 127.0.0.1 for server addresses.
• Port Conflicts: Check for port conflicts and adjust the port mapping in docker-compose.yaml if necessary.
• Nginx Configuration Errors: Verify that Nginx is forwarding requests to the correct service location.
• SSRF Proxy Configuration: Ensure SSRF_PROXY and HTTP_PROXY environment variables are correctly set to prevent SSRF attacks ^[3].

These steps should help you troubleshoot and resolve the connection issues with your local models in the intranet environment.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[laipz8200]

Thank you for the detailed logs. I've updated the code to use gpt2tokenizer as a fallback if tiktoken cannot be loaded.

[opsvr202]

感谢您提供详细的日志。我已更新代码，如果无法加载 tiktoken，则使用 gpt2tokenizer 作为后备。

谢谢，期待