Persistent 300s Read Timeout with OpenAI-API-Compatible provider

[caomuyunshui]

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 (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.

1. Is this request related to a challenge you're experiencing? Tell me about your story.

I am using Dify version 1.13.0. I have configured an LLM node using the "OpenAI-API-Compatible" model provider.
The Issue:
My backend model is running on Moore Threads S80 GPUs. Due to the hardware characteristics, the prefill stage (processing the prompt) takes an exceptionally long time. Consequently, I am constantly encountering the following timeout error:

error type: InvokeConnectionError
error message: [models] Connection Error, HTTPConnectionPool(host='192.168.12.174', port=8080): Read timed out. (read timeout=300)

What I've tried:
I have attempted to adjust the relevant environment variables to increase the timeout limit, but it seems ineffective. The system strictly enforces the 300-second timeout.

Question:
Is there a specific configuration file or parameter I am missing? How can I extend the timeout beyond 300s to accommodate the long prefill time?

2. Additional context or comments

No response

[dosubot[bot]]

Hi @caomuyunshui! I'm Dosu and I’m helping the dify team.

The 300-second timeout comes from Dify's plugin daemon architecture, which mediates all model provider invocations. The key environment variable is PLUGIN_DAEMON_TIMEOUT, which defaults to 300s in many deployments [1]. This timeout is applied to every HTTP request to the plugin daemon [2].

To fix this, add these environment variables to both your api and plugin_daemon services in your .env or docker-compose.yaml:

# Core timeout settings (must be kept in sync)
PLUGIN_DAEMON_TIMEOUT=1200
PLUGIN_MAX_EXECUTION_TIMEOUT=1800

# HTTP request timeouts
HTTP_REQUEST_MAX_CONNECT_TIMEOUT=3600
HTTP_REQUEST_MAX_READ_TIMEOUT=1800
HTTP_REQUEST_MAX_WRITE_TIMEOUT=1800

# Server timeouts
GUNICORN_TIMEOUT=600

# If using Nginx
NGINX_PROXY_READ_TIMEOUT=1800s
NGINX_PROXY_SEND_TIMEOUT=1800s

Then restart all containers:

docker-compose down && docker-compose up -d

Important notes:

• PLUGIN_DAEMON_TIMEOUT (API side) and PLUGIN_MAX_EXECUTION_TIMEOUT (daemon side) must be kept in sync [3].
• These variables are intentionally omitted from the default config files, so you need to add them manually [4].
• Dify has multiple independent timeout layers (plugin daemon, HTTP client, Nginx proxy, SSRF proxy), and each needs to be configured separately. If you're still hitting timeouts after setting the above, you may also need to adjust SSRF proxy settings [3].
• There's currently no way to set per-provider timeouts — these apply globally to all model providers.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}
   

[caomuyunshui]

I've configured it, but it still isn't working.

PLUGIN_PYTHON_ENV_INIT_TIMEOUT=1200

Plugin Daemon side timeout (configure to match the API side below)

PLUGIN_MAX_EXECUTION_TIMEOUT=6000

API side timeout (configure to match the Plugin Daemon side above)

PLUGIN_DAEMON_TIMEOUT=6000.0

HTTP_REQUEST_MAX_CONNECT_TIMEOUT=10
HTTP_REQUEST_MAX_READ_TIMEOUT=3000
HTTP_REQUEST_MAX_WRITE_TIMEOUT=3000

Proxy settings

NGINX_PROXY_READ_TIMEOUT=3600s
NGINX_PROXY_SEND_TIMEOUT=3600s

GUNICORN_TIMEOUT=3600

[caomuyunshui]

also

~/dify/docker$ cat .env |grep HTTP_REQUEST_MAX_CONNECT_TIMEOUT HTTP_REQUEST_MAX_CONNECT_TIMEOUT=3000