Docs Getting started

Getting started.

Three things to read before the first pipeline: the concepts Lima uses, the onboarding path an estate follows, and the thirty-minute first pipeline that shows the parts working together.

Concepts

ConceptWhat it is
ConnectionA registered way to reach one system, with credentials referenced from the vault and a tested endpoint.
CatalogThe inventory of tables, columns, definitions, owners and quality scores across every connected system; shared by questions and pipelines.
PipelineA sequence of operators on a canvas: connect, extract, transform, load, orchestrate, and where needed return.
RunOne execution of a pipeline, with a report: rows in, rows out, control totals, exceptions, deadline met or missed.
Run windowThe period in which a run must finish, with a deadline on the business calendar; a failed check pauses the run and the clock.
Exception recordA row that failed a check: kept, routed to an owner, replayed once fixed.
Access policyRole and attribute rules deciding who sees which rows and columns, applied at query time and at load time.
Write-backA result written into the system that owns it through its released interface; confirmed by that system; recorded in the ledger.

The onboarding path (22 days on the showcase estate)

Week 1: connectConnections registered and tested; the catalog built from the discovered schemas; the access policy applied from your identity provider's groups.
Week 2: rules and baselinesQuality rules written in plain language against live rows; KPI definitions agreed and the baseline read from your systems.
Week 3: pipelines or agentsThe first pipelines built on the canvas with checks on; or the first two governed agents proposing actions for approval.
Day 22: productionSchedules on the business calendar; the first reconciliation report; the first delta report scheduled.

A first pipeline in thirty minutes

Register a source and a targetPick a connector, reference the credentials from the vault, test the connection. A failed test names the cause.
Choose tables and a patternSelect tables from the discovered schema; choose full load for the first run.
Add a checkA completeness rule such as 'every item has a plant', written in plain language and previewed on live rows.
Run itWatch counts per step on the canvas; open the report: rows in, rows out, control totals, exceptions held.
Schedule itA run window with a deadline; alerts to the channel you choose.

Environments: development, test and production, with configuration promoted through reviewed changes in Git. The glossary defines every term used above: glossary.