Internal Engineering

This topic is for DataReadyAI engineers who build and ship the Nexa platform. It documents how the pieces fit together across dozens of repositories, how images are built and deployed, and how to run recurring operations. It is distinct from the customer-facing Deploy & Operate topic: that describes what the customer installs; this describes how we produce and wire it.

Nexa is customer-managed software. We ship container images and Infrastructure-as-Code; the customer’s platform team runs it in their own cloud (AWS or Azure) against their data platform (Databricks or Snowflake). Because every deployment is one cell of that 2×2 matrix, most internal pages absorb the variants into synced tabs rather than duplicating content — see the authoring guide.

Start here

Docs Authoring GuideHow this docs set is structured and how to add a page without breaking the build.

Platform ArchitectureEnd-to-end: web → backend → APIs → data platform → agents.

Cross-Repo MapWhich repository owns what, and how they connect.

Repository CatalogEvery product, API, IaC, and deployment repo, grouped by purpose.

CI/CDBuild, scan, sign, and publish images to ECR.

GitOps & DeploymentsArgoCD app-of-apps and image-updater write-back from nexa-deployments.

The two variant axes

Everything internal is shaped by two independent choices the customer makes at install time:

Axis	Options	Primary repos affected
Cloud	AWS, Azure	aws-iac vs nexa-azure-app-iac; ECR vs ACR; EKS vs AKS
Data platform	Databricks, Snowflake	nexa-databricks-api vs nexa-snowflake-api; agents/automations variants

See AWS vs Azure and Snowflake vs Databricks for how each axis changes the build and deploy.