How oracle-quickstart/oci-ai-blueprints works
Project type and purpose
• This is an Infrastructure-as-Code project built with Terraform.
• It provisions an Oracle Cloud Infrastructure (OCI) container platform (OKE) and then installs the “OCI AI Blueprints” platform on top of it—providing a no-code, opinionated solution for deploying and managing Generative AI workloads (inference, fine-tuning, autoscaling, multi-node, MIG, RDMA, etc.).
File structure
• Top-level directories:
• cluster_creation_terraform – Terraform code to provision a fresh OKE cluster (networking, security lists, data sources, variables, providers).
• oci_ai_blueprints_terraform – Terraform code to install the AI Blueprints platform (app-api, app-background workers, portal UI, config maps, database, ingress, RBAC, Helm chart for LWS CRDs, Corrino modules for cluster tools).
• docs & sample_blueprints – User docs and example blueprint JSON configurations.
• Patterns:
– A clear two-layer IaC architecture (cluster provisioning vs. platform deployment).
– Module-based Terraform (e.g. corrino modules for cluster-tools, networking, policies, vault).
– Kubernetes artifacts managed via Helm (lws Chart) and Terraform Kubernetes provider.
• Ancillary files: README, LICENSE, RELEASE_NOTES, version files.
Extract insights from the README
• Confirms the two-stage install: “Install AI Blueprints to an OKE cluster” then “Deploy Blueprints” (inference, fine-tuning, autoscaling, etc.).
• Lists key services: UI portal, API, background workers, monitoring, logging, auto-scaling.
Elements
• Terraform → OCI (resource creation)
• Terraform → Kubernetes (manifests & Helm)
• User → Portal/API → database & job queue
• API → LWS CRD → Inference Pods
• KEDA → Pod scaling
• Pods → Object Storage (model access)
• Prometheus → Pods & services → Grafana/Jaeger
• Infrastructure (blue)
• Kubernetes control plane & nodes (green)
• Microservices (orange)
• Data stores (purple)
• Observability (teal)