Chapter 1. Evaluate OpenShift Lightspeed capabilities

Use Red Hat OpenShift Lightspeed to fix and manage your OpenShift clusters. You can operate the virtual assistant by using plain English right inside the OpenShift web console.

Hardware and software requirements for OpenShift Lightspeed, including supported OpenShift Container Platform versions and CPU architectures.

OpenShift Container Platform clusters enable telemetry by default.

OpenShift Lightspeed supports Software as a Service (SaaS) and self-hosted large language model (LLM) providers that meet defined authentication requirements.

The LLM is a type of machine learning model that interprets and generates human-like language. When you use the LLM with a virtual assistant, the LLM can accurately interpret questions and offers helpful answers in a conversational manner. The OpenShift Lightspeed Service must have access to the LLM provider.

The Service does not provide the LLM for you, so you must configure the LLM before installing the OpenShift Lightspeed Operator.

The OpenShift Lightspeed Service can rely on the following SaaS LLM providers:

If you want to self-host a model, you can use Red Hat OpenShift AI or Red Hat Enterprise Linux AI as your model provider.

Red Hat OpenShift Lightspeed supports Federal Information Processing Standards (FIPS). You can run Red Hat OpenShift Lightspeed on OpenShift clusters that use FIPS mode.

FIPS is a set of publicly announced standards developed by the National Institute of Standards and Technology (NIST), a part of the U.S. Department of Commerce. The primary purpose of FIPS is to ensure the security and interoperability of computer systems used by U.S. federal government agencies and their associated contractors.

OpenShift Lightspeed works with OpenShift Container Platform clusters that use the x86_64 architecture.

OpenShift Lightspeed works in disconnected clusters without full internet access.

In a disconnected cluster, you must mirror the container images you need. For more help, see "Mirroring in disconnected environments" in the OpenShift Container Platform documentation.

OpenShift Lightspeed communicates directly with your configured large language model (LLM) provider endpoint. No Red Hat-managed cloud services or proxies are involved in the data path between OpenShift Lightspeed and the LLM provider during active chat interactions.

When you submit a prompt, the OpenShift Lightspeed Service pod on your local cluster sends the request directly to your configured LLM provider endpoint. Supporter providers include Azure AI Foundry, OpenAI, and self-hosted models served by OpenShift Container Platform AI. The LLM provider returns its response directly to the OpenShift Lightspeed Service pod. Red Hat infrastructure does not process, route, or store any prompt or response data during this exchange.

Separately, if telemetry is enabled, OpenShift Lightspeed sends conversation data to Red Hat by using the Red Hat Insights system. This telemetry transmission is independent of the prompt and response data flow and can be disabled.

OpenShift Lightspeed adds cluster and environment details to your messages. Then, it sends this data directly from the OpenShift Lightspeed Service pod on your cluster to the configured large language model (LLM) provider endpoint. No Red Hat-managed cloud services are involved in this data exchange.

OpenShift Lightspeed has limited ability to filter or hide the data you send to the LLM. Do not enter any information into the interface that you want to keep private from the LLM.

When you send transcripts or feedback to Red Hat, you agree that Red Hat can use the data to improve our Service. The transcript recording data uses the Red Hat Insights system. It follows the same security rules and access limits as that system. Telemetry and remote data collection can be completely disabled at your discretion.

To request deletion of your data, open a support ticket with Red Hat.

OpenShift Lightspeed sends your messages and cluster data through a redaction layer. It does this to clean the data before it goes to the LLM.

Do not enter anything into the OpenShift Lightspeed interface that you want to keep private from the LLM.

Conversation data, including user inputs, LLM outputs, tool call details, retrieval-augmented generation (RAG) context, and associated metadata, is only sent to Red Hat if telemetry is enabled. You can disable telemetry to stop all data collection by Red Hat.

The transcript recording data uses the Red Hat Insights system. It follows the same security rules and access limits as that system.

Remote Health Monitoring uses the Telemeter Client and Insights Operator to gather and report cluster information for Red Hat analysis and support.

You can learn how Red Hat collects data in the OpenShift Container Platform documentation. To stop sending chat transcripts or feedback, you must opt out of remote health monitoring. Follow the steps in the "About remote health monitoring" section of the OpenShift Container Platform documentation.

OpenShift Lightspeed stores interaction data locally in the OpenShift Lightspeed Service pod logs and in your cluster’s configured log aggregation system. Red Hat does not have access to this local data.

If telemetry is enabled, conversation data is sent to Red Hat through the Red Hat Insights system. This data includes user inputs, LLM outputs, tool call details, retrieval-augmented generation (RAG) context, and associated metadata, Red Hat retains data that is collected through telemetry until you submit a formal request for data deletion.

To request deletion of your telemetry data, open a support ticket with Red Hat. Include your cluster identifier and the scope of data that you want to be deleted.

You can disable telemetry at any time to stop all future data collection. Disabling telemetry does not affect data that Red Hat has already collected. To delete previously collected data, you must submit a separate deletion request.

Red Hat does not use customer prompt or response data to train or fine-tune AI models.

When telemetry is enabled, Red Hat uses the collected data exclusively to evaluate and refine internal system prompts and embedding parameters to improve OpenShift Lightspeed Service quality. Any interaction data that is retained for internal evaluation is heavily redacted and anonymized before use.

Red Hat does not train or tune its own AI models using customer interaction data, and does not share customer interaction data with third-party AI model providers for training purposes.