Provider setup fails with proxy: timeout errors

[alexeysn]

Hi team,

I'm using ragflow:v0.18.0, deployed in a Kubernetes cluster using the official Helm chart from this repository. Since there is no direct internet access, we configured environment variables for proxy:
HTTP_PROXY=http://
HTTPS_PROXY=http://
NO_PROXY=localhost,127.0.0.1,.svc,.googleapis.com

If run "curl https://generativelanguage.googleapis.com" from pod, returns a valid response, confirming that the proxy setup works correctly.

However, when I try to add a Gemini model provider, I receive the following errors in logs:
2025-05-19 13:07:38,742 ERROR 19
Fail to access embedding model(embedding-001) using this api key. Timeout of 60.0s exceeded, last exception: 503 failed to connect to all addresses; last error: UNKNOWN: ipv4:173.194.73.95:443: Failed to connect to remote host: Timeout occurred: FD Shutdown
Fail to access model(gemini-2.0-flash-thinking-exp-01-21) using this api key. ERROR: Unknown field for GenerationConfig: max_tokens
Fail to access embedding model(text-embedding-004) using this api key. Timeout of 60.0s exceeded, last exception: 503 failed to connect to all addresses; last error: UNKNOWN: ipv4:64.233.165.95:443: Failed to connect to remote host: Timeout occurred: FD Shutdown
NoneType: None
Also, I don't see any proxy access attempts in the logs — it seems like the requests are not being routed through the configured HTTP/HTTPS proxy.

[dosubot[bot]]

Hi @alexeysn! I'm Dosu, a bot here to help you solve bugs, answer questions, and navigate the world of contributing. I'm ready to assist you while we wait for a human maintainer. Let's tackle this issue together!

RAGFlow v0.18.0 may not be correctly handling proxy settings for model provider connections, which could be causing the timeout errors you're experiencing. Here are a few things to check and try:

1. Proxy Configuration: Ensure that the proxy settings are correctly configured in the service_conf.yaml.template file. You may need to update the base_url entry under user_default_llm to ensure that requests are routed through the proxy ^[1].

2. Baidu Plugin: If you are using the Baidu plugin, you might need to modify the ragflow/agent/component/baidu.py file within the Docker container to add proxy settings directly, as global Docker proxy settings might not be sufficient ^[2].

3. Timeout Settings: The default timeout settings for model provider connections are set to 3600 seconds in the ragflow_config.yaml file. Ensure that these settings are correctly applied and consider increasing the timeout if necessary ^[3].

4. Logging: RAGFlow logs proxy access attempts in /var/log/nginx/access.log. If you don't see any proxy access attempts, it might indicate that the requests are not being routed through the proxy. Double-check your proxy settings and ensure that the environment variables are correctly set in the Kubernetes deployment ^[3].

5. GenerationConfig Error: The error regarding the "Unknown field for GenerationConfig: max_tokens" suggests that there might be an issue with the configuration of the GenerationConfig. Ensure that the fields in the GenerateParam class are correctly set and that only valid fields with positive values are included ^[4].

If these steps do not resolve the issue, further investigation into the specific configuration and deployment setup may be necessary.

To continue talking to Dosu, mention @dosu.

---

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

[ZhenhangTung]

@alexeysn Hi. Sorry for the late reply. Will forward it to our eng team to take a deeper look.

[bhavik4it]

Hey guys,

I am getting the same. what is the solution to this ?

[yongtenglei]

Hi, @bhavik4it

What version are you using? Have you tried the latest one?

[ZhenhangTung]

@bhavik4it @alexeysn

Could you try our demo site to see if this error still exist?