Schema

Every Omnigraph repository starts with a schema. The schema is a single .pg file that declares what kinds of entities and relationships exist in your graph, what properties they carry, and what rules apply to the data.

What a .pg file looks like

node Person {
    name: String @key
    age:  I32?
}

node Company {
    name: String @key
}

edge Knows: Person -> Person {
    since: Date?
}

edge WorksAt: Person -> Company

The three declaration kinds

Every node and edge gets its own storage table. Interfaces are compile-time only.

What the schema drives

When Omnigraph opens a repository, it compiles _schema.pg into a catalog:

1. Typecheck queries — property references, traversal paths, and filter expressions are validated against the declared types.
2. Create and validate storage — one data table is created per node type and per edge type.
3. Enforce constraints — @range, @check, and edge cardinality annotations are checked at load and mutation time.
4. Build indices — @index and @key annotations cause BTree, inverted (full-text), or IVF-HNSW (vector) indices to be created.