ER (entity-relationship) diagrams

What is an ER diagram?

An entity-relationship (ER) diagram models the data a system stores. An entitybecomes a table (e.g. User, Order); its attributes become columns; and relationships describe how entities reference one another. It’s the blueprint a relational database is built from.

Keys

Primary key (PK)

A column (or set) that uniquely identifies each row — usually an id. Skeema marks it with a key icon and injects one automatically if the model forgets.

Foreign key (FK)

A column that references the primary key of another table, creating a relationship. Marked with a link icon.

Unique / nullable

Constraints shown as UQ and ? modifiers — whether a value must be unique, and whether it may be empty.

Cardinality & junctions

Relationships have a cardinality: one-to-one, one-to-many, or many-to-many. A many-to-many (e.g. students ↔ courses) can’t be stored directly — it needs a junction table (a join table) holding two foreign keys. Skeema detects junctions automatically (an entity with exactly two FKs and little else) and tags them with a JUNCTION badge, placing them lower in the layout so the parent tables read clearly.

Normalization (the short version)

Normalization is organizing columns so each fact lives in exactly one place, which prevents update anomalies. The practical rules Skeema’s health check looks for:

• •1NF — atomic columns; no repeating groups or comma-lists in a field.
• •2NF — non-key columns depend on the whole primary key, not part of it.
• •3NF — non-key columns depend only on the key, not on other non-key columns.

How Skeema generates ER diagrams

• •From a prompt — describe the domain; Skeema returns 5–12 entities with keys, enums, and inferred FKs.
• •From a database node — derive the schema directly from your architecture, carrying lineage.
• •From pasted schema — paste Prisma, PostgreSQL, or MySQL and a hybrid regex-plus-AI parser reconstructs the diagram.

Validation & code

Every ER board runs the schema-health check (9 rules, graded A–F) and can emit code through the codegen drawer — Prisma, PostgreSQL, MySQL, and Mongoose — deterministically, with no AI cost. See /docs/schema-health and /docs/code-generation.