openstack-knowledge

Composed OpenStack expertise for AI agents — service internals, deployment with kolla-ansible, day-two operations, upgrades, and an interactive learning harness.

This is the durable, iterable knowledge base behind the Learn OpenStack experience on stackers.network. It consolidates what used to be three separate repos (openstack-core, openstack-kolla, and the learning parts of common-skills) into one place that can be versioned and improved over time — and consumed three different ways.

Independent and community-maintained. Not affiliated with or endorsed by the OpenInfra Foundation. "OpenStack" is used under nominative fair use to describe what this covers.

Three ways to consume it

1. Clone it (offline / CLI-native). Point Claude Code, Codex, or opencode at this directory — they read CLAUDE.md / AGENTS.md and gain the whole knowledge base. Works with no website and survives anything happening to any hosted service.

   git clone https://github.com/stackers-network/openstack-knowledge
   # then, in the repo, just ask your agent: "Teach me OpenStack" / "Guide me through a deploy"

2. Over the web (zero install). Tell any agent:

   Read https://stackers.network/learn.md and teach me OpenStack.

   The site renders this repo into a /learn.md manifest your agent can follow, fetching individual skills on demand.

3. Browse it (humans). stackers.network/learn renders the knowledge as readable pages.

Three learning modes

Mode	Say…	What happens
orient	"Where do I start with OpenStack?"	Surveys what's here, recommends a path for your goal and level.
train	"Teach me OpenStack fresh."	Assesses your level, builds a curriculum from core/, teaches concepts then procedures with exercises.
guide	"Guide me through a deployment."	Assesses your environment, builds a plan, walks you through deploy/ step by step.

Layout

harness/        learning experience — orient / train / guide
core/           service knowledge — foundation + identity, compute, networking, storage, …
deploy/         kolla-ansible lifecycle — config-build, deploy, health-check, diagnose, day-two
operations/     cross-cutting ops — upgrades (incl. SLURP), backup-restore, capacity-planning
reference/      decision-guides, compatibility matrix, known-issues
manifest.yaml   machine-readable index of every skill + consumption + live-grounding config
CLAUDE.md       agent entry: identity, critical rules, routing table, workflows
AGENTS.md       portable entry pointer for other agents

Each skill is a directory whose README.md is the entry point; service skills add architecture.md, internals.md, and operations.md.

The live edge

The knowledge here is deliberately timeless. stackers.network adds the now: when the harness teaches a service, it can pull that project's weekly code-activity digest (e.g. stackers.network/projects/openstack-neutron/feed.xml) so a lesson on Neutron opens with what actually merged this week. Frozen fundamentals + live community pulse.

Status

Scaffold. The harness/, manifest.yaml, and agent-entry files are authored. The knowledge skill entries are laid out with structure + migration pointers — port the authoritative content from the prior openstack-core / openstack-kolla repos into each skill's README.md (+ architecture.md / internals.md / operations.md).

Contributing

Improvements to skills flow here, then ripple to every consumer (clones, the web manifest, and the rendered pages) on the next sync. Keep skills release-aware (2025.1 → 2026.1) and prefer concrete commands and configs over prose.