ocp-lightspeed-about

name	ocp-lightspeed-about
metadata	{"author":"rhoai3-coding-demo","version":"1.0.0","platform-family":"ocp","platform-baseline":"repo","ocp-baseline":"repo","skill-group":"OpenShift Platform"}
description	Use when documenting, reviewing, or explaining OpenShift Lightspeed concepts, architecture, AI assistant capabilities, and supported LLM providers from the official OpenShift Lightspeed 1.0 documentation. Do NOT use for installing (use ocp-lightspeed-install), configuring (use ocp-lightspeed-configure), operations (use ocp-lightspeed-operate), or troubleshooting (use ocp-lightspeed-troubleshoot).

OCP Lightspeed About

Use this skill to ground OpenShift Lightspeed conceptual guidance in the official Red Hat OpenShift Lightspeed 1.0 about guide for the active baseline in docs/PLATFORM_BASELINE.md.

Source Grounding

Read references/source-capture.md before using product behavior. Official Red Hat documentation is product authority.

What Is OpenShift Lightspeed

Red Hat OpenShift Lightspeed is a generative AI service that provides context-aware recommendations for OpenShift Container Platform. It runs as a virtual assistant inside the OpenShift web console, accepting plain English queries from developers and administrators.

The service answers questions using official OCP documentation. For products not fully covered by OCP docs (GitOps, Pipelines, Service Mesh 3.x, ACS, ACM, Quay, and others), the configured LLM generates answers directly.

Supported Interfaces

• REST API (/v1/): stable, supported interface for programmatic use.
• Gradio UI (/ui) and internal endpoints: development and debugging only; unstable and unsupported for programmatic use.

Supported LLM Providers

The service does not include an LLM. You must configure one before installing the operator.

SaaS providers:

• OpenAI
• Microsoft Azure OpenAI
• IBM watsonx

Self-hosted providers:

• Red Hat OpenShift AI (single-model serving platform, vLLM runtime, vLLM 0.8.4+)
• Red Hat Enterprise Linux AI (OpenAI API-compatible, vLLM 0.8.4+)

Both self-hosted options are OpenAI API-compatible. When running in a different cluster than the Lightspeed deployment, expose via route or secure endpoint.

Resource Requirements

Component	Min CPU	Min Memory	Max Memory
Application server	0.5 cores	1 GB	4 Gi
PostgreSQL database	0.3 cores	300 Mi	2 Gi
OCP web console	0.1 cores	50 Mi	100 Mi
Lightspeed operator	0.1 cores	64 Mi	256 Mi

Platform Requirements

• Architecture: x86_64
• FIPS mode supported
• Disconnected clusters supported (mirror images with oc mirror)

Data Handling and Telemetry

Lightspeed adds cluster and environment details to user messages before sending to the LLM. A redaction layer filters data before sharing or logging.

• Transcripts sent to Red Hat every two hours by default (via Red Hat Insights)
• Opt-in feedback collection (score, text, query, LLM response)
• Disable via OLSConfig CR (feedbackDisabled, transcriptsDisabled)
• Full opt-out requires disabling cluster remote health monitoring

OLSConfig CR

The only CR documented in the About guide:

apiVersion: ols.openshift.io/v1alpha1
kind: OLSConfig
metadata:
  name: cluster

Key data-collection fields under spec.ols.userDataCollection:

• feedbackDisabled: stops feedback collection
• transcriptsDisabled: stops transcript collection

Workflow

1. Confirm the active OpenShift baseline in docs/PLATFORM_BASELINE.md.
2. Read references/official-doc-extraction.md.
3. Identify the relevant concept (LLM providers, data handling, requirements).
4. For GitOps manifests involving OLSConfig, verify fields against the extraction before committing.
5. For live operations, use the repo environment guard.
6. Validate with references/source-capture.md boundaries.

Related Skills

• Use ocp-lightspeed-install for installing the operator and deploying the service.
• Use ocp-lightspeed-configure for OLSConfig CR configuration and LLM setup.
• Use ocp-lightspeed-operate for day-2 operations and administration.
• Use ocp-lightspeed-troubleshoot for diagnosing and resolving issues.
• Use rhoai-model-serving-platform for RHOAI single-model serving (self-hosted LLM).
• Use rhoai-model-deployment for deploying models on RHOAI (self-hosted LLM).

References

• references/source-capture.md
• references/official-doc-extraction.md