Chapter 1. Troubleshooting OpenShift Lightspeed
The following topics provide information to help troubleshoot OpenShift Lightspeed.
1.1. 502 Bad Gateway Errors in the interface Copy linkLink copied to clipboard!
If you try to start using OpenShift Lightspeed before the internal components have stabilized, 502 Bad Gateway errors can occur. While all the pods might be running, there can be other console and internal OpenShift Container Platform system components that are not ready.
Wait a few minutes and try using OpenShift Lightspeed again.
1.2. Operator Missing from the Operator Hub list Copy linkLink copied to clipboard!
If you are not running your OpenShift Container Platform cluster on an x86 architecture, for example OpenShift Local on M1 Mac, you will not see the OpenShift Lightspeed Operator in Operator Hub due to the way the Operator Hub filters architectures.
OpenShift Lightspeed is only supported on x86 architecture.
1.3. Thinking model generates delineator prompt Copy linkLink copied to clipboard!
In some cases, thinking or reasoning models produce output that has delineators, such as THOUGHT
or reasoning
, to separate the thinking process from the model output that the OpenShift Lightspeed Service uses to answer your question.
The OpenShift Lightspeed Service does not configure the delineator behavior or produce the delineators in the output. The delineator feature is a model configuration setting. Typically, you can disable the feature using one of the following methods:
-
Append your question prompt with the specific keyword that the model requires to disable the delineator feature, for example,
/nothink
. For more information, see the documentation for the model you are using. - Modify the inference server configuration settings to disable the delineator feature. For more information, see the documentation for the inference server or for the model you are using.