Forum

Hello, I have investigated the issue for you, and it shares the same root cause as SUP-12036:

Cause nousresearch/hermes-agent:latest requires loading a large number of plugins upon startup (logs show minecraft-modpack-server, comfyui, godmode, etc., are being bundled, totaling 87 plugins). The total startup time exceeds the default Zeabur Startup Probe timeout, causing the container to be marked as failed and restarted by the probe before the Hermes Gateway finishes starting, creating a loop.

Additionally, a configuration conflict is exacerbating the issue:

• The service currently only exposes port 8080
• However, the environment variables are set to API_SERVER_PORT=8642 and API_SERVER_ENABLED=false

This means that even if the probe waits long enough, there is no endpoint listening on 8080, so the probe will always fail.

Solution Currently, Zeabur does not allow users to manually adjust the Startup Probe timeout. Please follow these steps:

1. Go to the "Settings" tab of the service, find "Health Check", and disable it.
2. In "Environment Variables", change API_SERVER_ENABLED to true.
3. Change the service's external port to 8642 (matching API_SERVER_PORT), or change API_SERVER_PORT to 8080.
4. Restart the service and wait for Hermes to finish loading the plugins.

If it still fails to start after disabling the Health Check, please reply with the full startup logs, and we will assist you further.

If you have any other questions, please open a new post. Thank you!

It is working properly now, thank you.