How to read this
runs once for all of Gemini
runs once per client (×N)
defined, but no client enables it
notebook run job condition for-each
Scroll to zoom · drag to pan · click to dive in · Esc to go up
Gemini Data Team — Knowledge Map
Everything here is generated from the databricks.yml and resources/ of
db-ingestion, db-customer-ingestions and data-science. Nothing is hand-maintained.
- Zoom with the wheel/trackpad, pan by dragging. Zoom in on a job to see its task DAG.
- Click a repo, job or task to focus it and open the detail panel. Click a job's ⇢ links to jump to the job it calls.
- Pick a client (top bar or the chips) to see exactly which workloads run for it. Shared, gemini-wide jobs stay lit — they run for everyone.
- Keys: / search · Esc up one level / close panel · Home fit all.
- Deployment tab (top bar): how code ships in each repo — branches, CI, targets, workspaces and
Aurora. Infra facts and "watch out" notes are curated in
extract/static/infra.yml. - Domains tab: the handover handbook — the domains the data department runs, one page each
(written in
site/handbook/). Job names on a page open that job on the map.
Refresh the data with make extract (or just make serve); make view serves
the committed data without needing the data repos cloned.