| name | rhoai-ai-pipelines |
| metadata | {"author":"rhoai3-demo","version":"1.0.0","platform-family":"rhoai","platform-baseline":"repo","ocp-baseline":"repo","skill-group":"RHOAI Platform"} |
| description | Use when documenting, reviewing, or operating Red Hat OpenShift AI AI Pipelines from the Working with AI pipelines guide: pipeline server configuration, S3-compatible artifact storage, default versus external MySQL/MariaDB metadata databases, Kubernetes API pipeline definition storage, KFP 2.0 SDK compilation, Pipeline/PipelineVersion custom resources, import/version lifecycle, caching, experiments, runs, workspaces, logs, Elyra, DSPA troubleshooting, and reusable component pipeline lifecycle. Do NOT use for repo KFP component code edits (use rhoai-kfp-pipeline-authoring), project/workbench lifecycle, IDE usage outside pipelines, S3 notebook operations, certificate trust outside pipeline context, Spark applications, or live cluster changes without the OpenShift safety guard.
|
RHOAI AI Pipelines
Use this skill for Red Hat OpenShift AI AI Pipelines product workflows on the
active product baseline in docs/PLATFORM_BASELINE.md.
Source Grounding
Read references/source-capture.md before using product workflow details.
Official Red Hat documentation is product authority. This skill adapts the
official Working with AI pipelines guide to this repo's GitOps and demo
workflow model.
Scope
This skill covers:
- pipeline concepts: pipeline server, pipeline, pipeline version, experiment,
artifact, task execution, and run
- pipeline server configuration with S3-compatible object storage and database
choices
- external MySQL/MariaDB and Amazon RDS certificate trust handoff
- KFP 2.0 SDK pipeline compilation to IR YAML
- Kubernetes API storage,
Pipeline, PipelineVersion, and GitOps-aligned
pipeline definition management
- KFP SDK authentication to an OpenShift AI pipeline server
- pipeline import, delete, version upload/delete/view/download lifecycle
- caching behavior and cache controls at task, run, compile, and server levels
- experiments, artifacts, task executions, comparisons, archive/restore/delete
- active, scheduled, archived, duplicated, stopped, restored, and deleted runs
- pipeline run workspace support and external artifact copy pattern
- pipeline step logs and download behavior
- Elyra JupyterLab pipeline editor runtime configuration, run, and export
- DSPA component error interpretation and troubleshooting handoff
- product-side lifecycle for modular pipelines assembled from reusable
components, including import, versioning, experiments, runs, logs, and
artifact review
Use other skills for adjacent work:
rhoai-kfp-pipeline-authoring for editing repo KFP Python code, components,
runner scripts, and demo-specific pipeline implementation standards
rhoai-project-workflows for project, workbench, connection, and cluster
storage lifecycle
rhoai-data-science-ide-workflows for non-pipeline JupyterLab and
code-server IDE workflows
rhoai-s3-object-storage-data for notebook object-storage data operations
rhoai-automl for AutoML optimization runs, leaderboard evaluation,
AutoGluon pipeline naming, saved notebooks, and AutoGluon serving handoff
rhoai-autorag for AutoRAG optimization runs, imported AutoRAG pipeline
naming, leaderboard review, and generated indexing/inference notebooks
rhoai-model-customization-training for Docling, SDG Hub, Training Hub, and
end-to-end model customization pipeline patterns
rhoai-enterprise-rag for deciding whether a RAG stage should use Docling
and KFP automation for unstructured corpus ingestion
rhoai-evaluation for EvalHub risk assessment KFP orchestration and
official LM-Eval or EvalHub evaluation workflows
rhoai-certificate-management for DSCI trusted CA bundle changes
rhoai-storage-classes for OpenShift AI storage class administration
rhoai-model-evaluation for evaluation-specific pipeline evidence and
metrics patterns
rhoai-kubeflow-spark-operator for Spark data processing applications that
run through SparkApplication resources rather than AI Pipelines
Demo Policy
For this repo:
- Prefer Kubernetes API pipeline definition storage when the demo needs
reviewed GitOps desired state for pipeline definitions and versions.
- Keep pipeline server configuration in GitOps only after fields are verified
against official docs or active CRD schema.
- Treat the default on-cluster database as development/test only. For
production-positioned pipeline workloads, document the external MySQL or
MariaDB requirement and CA trust path.
- Use S3-compatible object storage for pipeline artifacts and keep credentials
project-scoped and out of Git.
- Do not pass OpenShift access tokens as literal command arguments. Use
environment variables or secure prompts.
- Use exact KFP SDK and Python prerequisites from the active official docs
before compiling or authenticating pipelines.
- Use caching intentionally. Disable it only for tasks or runs that require
deterministic re-execution, debugging, or frequently changing inputs.
- Treat archive as retention and delete as destructive. Require explicit
confirmation for deleting pipeline servers, pipelines, versions, archived
experiments, and archived runs.
- Use Elyra only with supported JupyterLab workbench images. Do not claim Elyra
support for code-server, RStudio, Minimal Python, or CUDA-based workbenches.
- If an official example uses placeholder or mutable model artifacts, replace
them with verified demo artifacts before committing demo content.
- For Docling data-processing automation, use the official model customization
data-preparation chapter plus
rhoai-model-customization-training for the
example source. This skill governs the pipeline server, runs, artifacts,
logs, versions, and dashboard lifecycle once the Docling pipeline exists.
- When a Red Hat Developer article or component catalog suggests reusable KFP
components, treat it as implementation guidance. Verify the pipeline server,
storage, run, experiment, artifact, and log behavior through this skill, then
route component code selection or authoring to
rhoai-kfp-pipeline-authoring.
- Prefer small validation runs before scaling pipeline workloads to full data
sets or expensive GPU-backed steps.
Workflow
- Confirm the active baseline in
docs/PLATFORM_BASELINE.md.
- Read
references/source-capture.md and
references/official-doc-extraction.md.
- Decide whether the task is:
- pipeline server setup or deletion
- pipeline definition, import, version, or GitOps/Kubernetes API storage
- KFP SDK compilation or authentication
- caching configuration
- experiment, artifact, or run management
- scheduling and duplicate-run workflow
- workspace or external artifact use
- logs, Elyra, or DSPA troubleshooting
- reusable component catalog adoption and pipeline run lifecycle
- Use
examples/ai-pipelines-patterns.md for focused review patterns.
- For repo KFP code edits, pair with
rhoai-kfp-pipeline-authoring.
- For live cluster work, follow the OpenShift safety guard in
AGENTS.md.
- Validate with
references/validation-checklist.md.
References
references/source-capture.md
references/official-doc-extraction.md
references/validation-checklist.md
examples/ai-pipelines-patterns.md