8.0. Repository

What is inside the repository?

Everything the course produces lives in a single repository, MLOps-Courses/agentops-open-course. It has three payloads — the course content, the reference agent, and the infrastructure — plus the tooling that binds them:

agentops-open-course/
├── README.md            # for humans — start here
├── AGENTS.md            # for AI coding agents (dogfoods the AGENTS.md standard)
├── CONTRIBUTING.md       # setup, quality gates, and pull-request expectations
├── CODE_OF_CONDUCT.md    # community participation standards
├── SECURITY.md           # private vulnerability-reporting process
├── CITATION.cff          # machine-readable citation metadata
├── CHANGELOG.md           # Keep a Changelog release history
├── LICENSE               # MIT — software, scripts, workflows, and configuration
├── mkdocs.yml           # Zensical config: theme, navigation, extensions
├── pyproject.toml       # root project: version + the pinned Zensical build dependency
├── mise.toml            # task runner: install/serve/build/format/check/test/scan
├── lefthook.yml         # git hooks → mise run tasks
├── dprint.json          # formatting for JSON, Markdown, TOML, YAML
├── .github/              # workflows plus issue and pull-request templates
├── docs/                # ← THE COURSE (this site): CC-BY-4.0 prose + license text
├── agents/              # ← THE REFERENCE AGENT: the Python project + shared data (MIT)
│   ├── python/          # ADK 2.0 Python project (uv, ruff, ty, pytest)
│   └── data/            # the bundled Ops Copilot dataset (incidents.db, sql/, runbooks/, skills/, logs/)
└── infra/               # agentgateway host/k3d/GKE, kagent, MLflow, OTel, OpenTofu (MIT)

docs/ mirrors the sibling MLOps Coding Course: one folder per phase, N.M section pages, and an index.md per folder, all written as FAQ Markdown.

Why keep the course, the agent, and the infra together?

Because they evolve as one. A monorepo keeps the prose, the code it documents, and the deployment artifacts in lockstep — a single clone, a single set of tasks, one version history. When Chapter 3 teaches a tool, the exact tool lives in agents/; when Chapter 6 deploys the agent, the exact manifests live in infra/. Nothing in the course points at a snippet you cannot open and run.

Why split the agent into python and data?

The reference agent — the DevOps "Ops Copilot" — is implemented as a Python project, split from the data it reads:

• agents/python is a self-contained project, with its own mise.toml and the standard task vocabulary (install, format, check, test).
• agents/data is separate: immutable SQLite/Markdown/log/skill seed input. The runtime copies SQLite into writable state; Kubernetes agent and MCP processes share one PVC so approved writes and later reads stay coherent. Rebuild the seed with cd agents/data && mise run build.

Course pages show short executable excerpts; the complete Python and infrastructure sources remain canonical.

README.md or AGENTS.md — which do I read?

Both exist on purpose, for two different audiences:

• README.md is for humans: the pitch, the feature list, the layout, and the install commands.
• AGENTS.md is for AI coding agents: the same repository described as machine-actionable rules — the layout, the mise run commands, the conventions, and the definition of done.

The repository practises what it teaches. Compatible assistants can read one shared file, and maintainers review their changes through the same mise run gates as human contributions.

Where are the contributor and security policies?

GitHub surfaces the repository's community files automatically:

• CONTRIBUTING.md gives contributors one setup path and the exact local quality gate.
• CODE_OF_CONDUCT.md defines participation and enforcement standards.
• SECURITY.md routes vulnerabilities away from public issues.
• CITATION.cff lets GitHub and reference managers generate a citation for the course.
• .github/ISSUE_TEMPLATE/ and .github/PULL_REQUEST_TEMPLATE.md ask for reproducible evidence and the same mise run validation used by CI.

What is the AGENTS.md standard?

AGENTS.md is a plain-Markdown file at the repository root that gives AI coding agents the instructions and context they need to work in a project. It is to agents what README.md is to humans: a predictable, agent-first entry point that any tool can look for and read.

The rest of this chapter walks through the other things that turn a repository into a shareable, sustainable project: its license, its releases, reusable templates, its documentation pipeline, the contribution workflow, and the AAIF it supports.