Private AI Platform Kit

This project is a Kubernetes reference implementation for a governed LLM gateway, retrieval service, and coding-agent workspaces. It has two deployment paths:

• a local kind cluster with Ollama for development and evaluation;
• a customer-cluster template with Ollama, vLLM, Qdrant, and agent workspaces.

The repository contains service code, Helm charts, Argo CD applications, policy, tests, and runbooks. It does not provision a cluster or replace the operator's identity, secrets, ingress, observability, or backup systems.

Maturity

Current release v0.28.0: reference implementation and customer lab. Treat the customer values as a template. Production use requires customer integration, capacity tests, backup and restore evidence, and the strict validation gate.

Start here

• Local quickstart explains the downloads, side effects, and completion checks for the local lab.
• Feature inventory lists the implemented API and platform features, including defaults and exclusions.
• Architecture describes the local and customer profiles and the restricted-egress tenant example.
• Decision guide covers fit, poor-fit cases, and the work a customer still owns.
• Production readiness maps platform controls to their validation commands.

Operating and reviewing the project

• Getting started collects the validation, local-cluster, eval, and evidence commands.
• Runbooks cover deployment, incidents, upgrades, data stores, and release checks.
• Security overview states the important defaults and links to the threat model.
• Release verification covers published images, charts, signatures, and checksums.
• Scope and non-goals defines the supported boundary.

The source repository also has a compact documentation map for browsing on GitHub.