Belt Documentation

Belt is a visual SQL-first ELT platform for designing, running, and monitoring data pipelines.

The platform was created to remove routine work from implementing common data loading patterns while preserving flexibility. Its goal is to make the creation of data loading flows as fast and simple as possible, and to allow analysts to build pipelines without programming skills or server command-line experience. Belt helps reduce the gap where analysts have to wait for data engineers to prepare and deliver data.

Belt is not just an executor and not a classic orchestrator that runs a set of pre-written scripts. At its core, Belt is based on a structured model: connections, flows, nodes, SQL queries, relationships between them, and loading modes. Each loaded data package also receives technical metadata that makes it possible to track its origin and manage loading results.

Because of this, Belt not only loads data, but also stores extended information about the process itself. The platform knows where the data comes from, which flows and targets it passes through, where it is loaded, which loading mode was used, and how objects are related to each other. This makes it possible to show lineage, control execution, analyze run history, manage loaded batch/request data, and automatically generate documentation for existing pipelines.

The main idea of Belt is to turn data loading from a set of scattered scripts and tools into a managed visual model that can be designed, executed, analyzed, documented, and maintained in one place.

This documentation describes the current live version of Belt.