Repositories

Fifteen repositories are public under github.com/datahub-local. Five are the platform and carry the work. The other ten are support artifacts: a chart, an image, or an example the platform consumes.

The project catalogue is organised by use case. This page is organised by repository, so you can go from a question to the place that answers it.

The platform

Repository	Owns	Read first
`datahub-local`	The published documentation site (MkDocs). Project pages, architecture, lessons learned.	`mkdocs.yml`, `docs/`
`datahub-local-core`	The helmfile releases ArgoCD applies: every service, its values, its patches.	`releases/<namespace>/helmfile.yaml.gotmpl`, `releases/<namespace>/values/`
`datahub-local-ai`	AI workflows and the data pipelines: n8n exports and prompts, Sympozium ensembles, dbt/dlt/Airflow, Superset bundles, adapter images.	`AGENTS.md`, then the deep dive for the path you touch
`datahub-local-ai-mcp`	The MCP servers agents call: `homelab_facts`, `semantic`, `semantic-finance`, `utils`, `opencode`.	`README.md`, `servers/<name>/`
`datahub-local-bootstrap`	Ansible that brings a machine up as a cluster node. Owner only.	`README.md`, `playbook/`

Where to go with a question

"How is this deployed, and why does this Pod fail?" datahub-local-core. The values file for that release is the answer; the chart is upstream.
"Why did this automation fail?" datahub-local-ai, under agents/n8n/.
"What does this metric mean, and what SQL computes it?" datahub-local-ai-mcp (semantic) with the registry in datahub-local-ai/workflows/dbt/semantic/.
"Add or fix a dashboard." datahub-local-ai/workflows/superset/.
"Change a page or a diagram." datahub-local.
"Change what an agent reports or when it runs." datahub-local-ai/agents/sympozium/.
"Add or fix an MCP tool." datahub-local-ai-mcp.
"Why does a host behave like this?" datahub-local-bootstrap for the role, read only. The running host itself is the owner's machine.
"How do I describe the platform?" datahub-local, which is the published text.

Support repositories

Nothing routine lands in these. A task that names one is the only reason to open it. Several are archived, and GitHub marks them on the repository page.

Repository	What it carries
`garage-helm`	Helm chart for Garage, the S3-compatible object store the lakehouse writes to.
`spark-apps-helm`	Helm chart for Apache Spark applications on Kubernetes.
`node-exporter-textfiles`	Container image for textfile exporter scripts that `node_exporter` collects.
`ollama-metrics`	Metrics collector and proxy that exposes Prometheus metrics for Ollama.
`servarr`	Helm chart for a Servarr media stack.
`Byparr`	FlareSolverr drop-in replacement for browser-based scraping.
`dagster-celery-k8s-oci`	Dagster image for Kubernetes and Celery deployments on ARM.
`dagster-user-code-example`	Example user-code repository for the Dagster images.
`container-helm-version-updater`	Action that updates container image and Helm chart versions.
`docker-lineageos`	Docker-based build environment for LineageOS.

Not public

datahub-local-secrets and datahub-local-private-config are private and out of scope. Nothing here links to them, quotes them, or describes their values. If a task needs a value that lives there, ask for that one setting.

The org keeps other private repositories for personal configuration. They are out of scope for the same reason.

The agent skill repo-map in datahub-local-ai carries the same table. Change both in the same pass.