Entity schema

An entity is a record type Chef stores for your application. Define it in schema.graphql; each entity needs a unique id: ID!. The quickstart creates this block summary:

type BlockSummary {
  id: ID!
  height: Int! @index
  hash: String!
  time: BigInt!
  transactionCount: Int!
}

Scalar	Handler / GraphQL representation
ID, String	String
Boolean	Boolean
Int	Bounded integer
BigInt	JavaScript bigint in handlers; decimal string in GraphQL

Use ! for required fields. Chef 0.6.0 supports scalar fields, bounded lists and entity relationships. Use these features in your own 0.6.0 project; the hosted Playground has its own demo schema.

Lists and relationships

type Transaction {
  id: ID!
  outputs: [TransactionOutput!]! @derivedFrom(field: "transaction")
}

type TransactionOutput {
  id: ID!
  transaction: Transaction! @index
  valueZat: BigInt!
  addresses: [String!]!
}

Store the reference as its target ID in the handler: transaction: txid. Generated handler types use a string for references. Derived fields such as outputs are computed at query time and cannot be written by handlers. The inverse must be a single entity reference with @index; it counts toward the existing project index limit.

Scalar and stored-reference lists contain at most 100 items; nested lists are unsupported. BigInt list items preserve exact integer strings in PostgreSQL/GraphQL and become bigint in handlers. Item and container nullability follow the declared GraphQL type.

query Breakdown($txid: ID!) {
  TransactionById(id: $txid) {
    outputs(first: 100) { valueZat addresses transaction { id } }
  }
}

Inverse lists accept first: 1..100 (default 25). If more children exist, the query fails with RELATION_LIMIT_EXCEEDED and QUERY_CHILD_LIST. Read every child through TransactionOutputList(where: { transaction: $txid }) using normal cursor pagination. Excess children are never silently omitted. Nested reads share the root query's repeatable-read snapshot and readiness gate, with a 500-unit request budget for reference lookups and inverse-list work. Existing depth, field and response-size bounds still apply. Filter lists through their child entities; scalar-list equality filters are not supported.

References do not impose database foreign keys or cascade deletes. A missing optional target returns null; a required missing target fails according to GraphQL non-null rules. Schema changes require a fresh namespace and replay. Existing scalar-only schema fingerprints are preserved.

The transaction breakdown example combines these relationships with previous-output resolution.

Generate types

In a starter project:

npm run codegen
npm run typecheck

Use stable IDs and exact integer values. Optional equality indexes use @index on supported fields.

Notes: To change an existing indexer's schema, use a new project and replay its data. Regenerate types from the schema rather than editing them by hand.