Private AI Code Assistant on Red Hat OpenShift

Copy link

Deploy a private self-hosted AI coding assistant on OpenShift so developers get AI-powered IDEs while no code leaves your environment

Table of Contents

Copy link

1. Detailed description
   • Architecture diagrams
2. Requirements
   • Minimum hardware requirements
   • Minimum software requirements
   • Required user permissions
3. Deploy
   • Option 1: ROSA (AWS)
   • Option 2: ARO (Azure)
   • Option 3: Existing OpenShift
   • Validating the deployment
   • Delete
4. Documentation
5. References
6. Tags

Detailed description

Copy link

The challenge

Copy link

Teams want IDE-integrated AI coding help without sending proprietary source to public model APIs. They also need governance: who can call the model, which model is served, and where data resides.

Our solution

Copy link

This quickstart deploys:

• OpenShift Dev Spaces workspaces with OpenCode (Continue, Cline, Roo Code optional)
• On-cluster inference with vLLM behind llm-d intelligent routing — default model Qwen/Qwen3.6-35B-A3B-FP8 (you can change it to any open-source model you want)
• An optional RHCL AI Gateway front door with per-developer API keys
• GitOps Helm charts (ROSA/ARO) or Helm-only install on an existing cluster

Features

Copy link

• Source and inference stay inside your OpenShift network boundary when configured that way
• OpenAI-compatible /v1 API for IDE extensions
• Prefix-cache–aware routing via llm-d Endpoint Picker (EPP)
• Multi-developer namespaces with ConfigMap-prewired extensions
• Optional observability (Grafana; Langfuse/OTel), guardrails, and MCP

Solution stack

Copy link

Architecture diagrams

Copy link

Developers → Dev Spaces → RHCL AI Gateway → llm-d → EPP → vLLM on NVIDIA GPU.

Full diagrams: docs/architecture.md.

Requirements

Copy link

Minimum hardware requirements

Copy link

• GPU: 1× NVIDIA GPU with enough VRAM for the model at FP8 (e.g. L40S 48 GB, A100 80 GB, or H100 NVL)
• Storage: PVC for model cache, typically 50–100+ Gi
• CPU / memory: enough for Dev Spaces workspaces plus platform operators (size for team concurrency)

Details: docs/requirements.md, docs/benchmarks.md.

Minimum software requirements

Copy link

Pick one path.

All paths

• OpenShift 4.20+
• Red Hat OpenShift AI 3.4 (3.3 minimum)
• OpenShift Dev Spaces
• NVIDIA GPU Operator + NFD
• Hugging Face token
• oc CLI

Existing OpenShift — operators above already installed, plus RHCL, helm v3. See deploy_existing_openshift/README.md.

ROSA — Terraform >= 1.4.6, AWS CLI, rosa CLI, ROSA entitlement. See PCA_Deployment_ROSA/README.md.

ARO — Terraform >= 1.4.6, Azure CLI, ARO + GPU quota. See PCA_Deployment_ARO/README.md.

Required user permissions

Copy link

• ROSA / ARO: Cloud admin to create the cluster, then OpenShift cluster-admin to install operators and sync GitOps.
• Existing OpenShift: OpenShift cluster-admin to deploy AI serving, DevSpaces, and (optional) the demo IDP.

Deploy

Copy link

Choose one path. Do not mix ArgoCD GitOps (ROSA/ARO) with the existing-OpenShift Helm flow on the same cluster unless you know how ownership overlaps.

Option 1: ROSA (AWS)

Copy link

Terraform provisions ROSA HCP and a GPU pool; ArgoCD syncs charts/ with values-rosa.yaml.

→ PCA_Deployment_ROSA/README.md

Option 2: ARO (Azure)

Copy link

Terraform provisions ARO and a GPU MachineSet; ArgoCD syncs charts/ with values-aro.yaml.

→ PCA_Deployment_ARO/README.md

Option 3: Existing OpenShift

Copy link

Cluster already has RHOAI, GPU Operator, and Dev Spaces. Install RHCL first, then:

# AI serving once
make ai-serving-deploy-existing-openshift HF_TOKEN=hf_xxx

# Each developer workspace
make devspace-deploy-existing-openshift DEV_NAMESPACE=<dev-ns>
# 2nd+ developers: HELM_ARGS='--set devspacesGlobalConfig.enabled=false'

→ deploy_existing_openshift/README.md

Validating the deployment

Copy link

After deploy, run cluster smoke tests (not CI):

make smoke
make smoke AI_NAMESPACE=ai-serving DEV_NAMESPACE=<dev-ns>

→ tests/cluster-smoke/README.md

Delete

Copy link

Existing OpenShift — remove with make:

make devspace-undeploy-existing-openshift DEV_NAMESPACE=<dev-ns>
make ai-serving-undeploy-existing-openshift

ARO — follow Destroying the Cluster.

ROSA — follow Destroying the Cluster.

Documentation

Copy link

References

Copy link

• Red Hat OpenShift AI
• OpenShift Dev Spaces
• llm-d
• vLLM
• Kuadrant / RHCL