Projects
The platform is not the point. What it produces is. This catalogue lists the fourteen use cases built on DataHub.local across four areas. Every card opens a full case study with the problem, approach, technologies, outcome, and source.
Data & Analytics
-
Ingestion, Iceberg storage in Garage, a Polaris catalog, Trino federation, dbt transformations, and Superset dashboards.
-
Supermarket invoices turned into a structured Iceberg dataset and a weekly spend digest, described by method and aggregates only.
-
A private analytics pipeline over personal documents with a governed SQL surface and no source rows published.
-
A catalogued, versioned SQL surface over the lakehouse so agents and BI tools query the same governed definitions.
-
dbt tests, explicit model lineage, and Iceberg snapshots applied to the lakehouse tables.
AI & Agents
-
Narrow, scheduled agents that investigate alerts and trends through bounded MCP tools and report in a fixed format.
-
Read-only Model Context Protocol servers exposing Prometheus, Loki, Kubernetes, Trino, Polaris, and Garage facts to agents safely.
-
A natural-language front door over cluster and repository state for operational questions.
-
Versioned, LLM-powered workflows for content drafting, diagram generation, and notification pipelines.
-
:material-gpu: Local inference + GPU node
A GPU worker serving a local model through Ollama for private, offline inference and agent workloads.
Platform & Operations
-
A seven-node heterogeneous Kubernetes cluster assembled, provisioned, and operated from Git.
-
ArgoCD reconciliation, layered repositories, quality gates, and automated dependency updates.
-
Metrics, logs, dashboards, and enriched alerting, plus a proactive AI monitoring loop above the rule-based one.
Community
-
Five published charts and exporters: Spark apps, Garage, Servarr, node-exporter textfiles, and Ollama metrics.