Core Concepts & Vocabulary

A Target Schema represents the “Source of Truth” for your downstream databases. It defines what valid, acceptable data must look like. A schema contains:

• Field Definitions: Expected column names and data types (e.g., Customer_Name as a String, Onboarding_Date as a Date).
• Validation Rules: Hard constraints attached to columns (e.g., unique constraints, email formatting, numerical ranges).
• AI Rules: Plain-English validation prompts that the AI compiles into execution rules.

A Project is the parent container for an automated data pipeline. It binds:

1. 1An Input Source (where the data originates).
2. 2An Associated Target Schema (which defines the rules).
3. 3An AI Mapping Recipe (how the input is cleaned and mapped).
4. 4An Active Destination (where clean data is written).

A Data Portal is a client-facing or vendor-facing ingestion endpoint. It allows external parties to upload files directly into an active project. Data Portals can be fully styled with your company’s branding, letting clients upload messy CSVs or PDFs and correct high-level errors under the guidance of a friendly AI Assistant.

The AI Assistant is your conversational companion, embedded directly within your workspace panels. It guides you through the planning phase, explains why validation rules failed in plain English, and executes bulk corrections across your dataset in response to natural-language instructions (e.g., “Capitalize the first letters of all names”).

Lineage is Elvity’s universal auditing mechanism. It tracks the exact lifecycle of every data point. By selecting any cell in the smart grid, you can instantly see if a value was mapped directly, calculated via a mathematical formula, edited manually by a human, or extracted via OCR from a multi-page document.