Installation Overview
CAIPE ships as a set of containerized services deployable via Helm or Docker Compose. Pick the path that matches your environment:
| Environment | Best for | Guide |
|---|---|---|
| Docker Compose | Local development and quick evaluation | Docker Compose → |
| KinD | Local Kubernetes cluster (CI / laptop) | KinD → |
| Helm (production) | Any Kubernetes cluster — EKS, GKE, AKS, on-prem | Helm → |
| EKS | AWS-managed Kubernetes | EKS → |
Prerequisites
- LLM provider — any OpenAI-compatible endpoint (Anthropic Claude, OpenAI, Google Vertex AI Gemini, local Ollama, LiteLLM proxy, etc.)
- Kubernetes 1.27+ (for Helm / KinD / EKS paths) or Docker Compose v2 (for local path)
kubectl,helm3.x, and optionallykindinstalled on your workstation
What gets deployed
A full CAIPE stack includes:
- supervisor-agent — orchestrates all sub-agent calls
- caipe-ui — web interface (React / Next.js)
- MongoDB — chat and agent-config persistence
- dynamic-agents — runtime for user-built custom agents
- RAG stack (optional) — ingestors, vector store, graph RAG server
- Slack / Webex bots (optional)
The minimal install (supervisor + UI + MongoDB) is enough to run the pre-built agent fleet.
Quickest path
# Interactive setup script — detects Docker or Kubernetes automatically
bash <(curl -fsSL https://raw.githubusercontent.com/cnoe-io/ai-platform-engineering/main/setup-caipe.sh)
Or via Helm directly:
helm upgrade --install ai-platform-engineering \
oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
--version 0.4.8 -f your-values.yaml
See Getting Started → Quick Start for a full walkthrough.
Persistence
To enable durable conversation checkpoints and cross-session memory, see the Persistence guide — it covers Redis, PostgreSQL, and MongoDB options.
Helm Chart Reference
The Helm section contains the full values.yaml reference for every
sub-chart, including agent profiles, ingress, ExternalSecrets, and LLM configuration.
Individual chart references are in the sidebar under Chart Reference.