Skip to content

Pipelines

A pipeline is the executable data-transformation code Nexa generates from your Raw → Curated → Consumption mappings, then deploys and runs on your data platform. On Databricks these are DLT (Delta Live Tables) pipelines; on Snowflake they are the equivalent scheduled transformation objects. Use the Pipelines screen (Jobs & Pipeline in the sidebar → Pipelines) to monitor the pipelines Nexa manages, and use the mapping screens to (re)generate their code.

You don’t hand-write pipeline code in Nexa. It is generated from the mappings you define across the three layers:

Layer What you define What it generates
Raw (bronze) Sources and ingested tables Landing/ingestion into raw tables
Curated (silver) Raw → curated field mappings and cleaning rules Transform code into typed, standardized curated tables
Consumption (gold) Curated → consumption entity mappings and ETL logic Transform code into business entities and 360° views

The generation itself happens on the mapping and canvas screens, not on the Pipelines screen:

  • Define and refine mappings on the Curated Layer and Consumption Layer screens.
  • Author per-column ETL (extract, transform, load) and review generated SQL and PySpark on the Canvas.
  • Generate the full set of artifacts from the Curated Landing Zone using Generate all Schemas, Entities, Mappings, Code Pipelines and Deployment Artefacts.

Once generated and deployed, the pipeline appears in the Pipelines list as a Nexa-managed pipeline. The Pipelines screen is scoped to those pipelines (it queries with nexa_managed=true), so it shows what Nexa created and deploys — not every pipeline in your workspace.

The Pipelines list shows every Nexa-managed pipeline with its latest run state. Each row has:

Column Meaning
ID Platform pipeline identifier
Name Pipeline name (select it to open the detail panel)
Mode Pipeline mode, for example triggered or continuous
Status Latest update state, such as running, succeeded, failed, or no updates
Last Updated Relative time of the last update, for example 2h ago
Actions For a failed pipeline, a Fix with Nexa button that starts AI-assisted diagnosis

Filter the list with the Filter dropdown:

  • Status Filter — All Statuses, Running, Failed, Canceled, Succeeded, No Updates.
  • Sort By — Name, Updated At, Created At, or Last Update Time, each ascending or descending.

Use the search box to find a pipeline by ID or name. A pipeline whose last update failed is shown with its name in red.

Select a pipeline name to open its detail panel, which has two tabs.

  • Pipeline Details — Pipeline ID, Creator, Run as, Tags, Health, and the target definition: Catalog, Schema, Channel, and whether it runs Serverless. An external-link icon opens the pipeline in your data platform.
  • Updates — the update (run) history. Each entry shows the update ID, its state (a failed update is flagged), and its creation time. Select Load More to page through older updates.

Pipelines transform data; jobs orchestrate and schedule work, including running pipelines. The two are managed side by side under Jobs & Pipeline.

Pipelines are DLT pipelines. The detail panel exposes DLT-specific fields such as Channel and Serverless, and the external-link icon opens the pipeline in the Databricks workspace.