Search & Query Metadata

Sixb validates every query against your ontology. A property is not queryable just because it exists — you opt each field into query surfaces with query metadata, and you set a search profile on the object type for keyword and vector search. This page covers both. For the query API itself, see Querying Objects.

Property Query Metadata

Set query on a property to expose it to queries. searchable is the gate: every other flag (and weight) requires searchable: true on the same property, or validate() rejects the ontology.

import { defineObjectType, link, prop, stringEnum } from "@sixb/core/ontology"
import { Customer } from "./customer"
import { Project } from "./project"

export const Invoice = defineObjectType({
  id: "Invoice",
  name: "Invoice",
  properties: [
    prop("id", "string", { required: true, primary: true }),
    prop("number", "string", {
      required: true,
      query: { searchable: true, filterable: true, exact: true, sortable: true },
    }),
    prop("amount", "double", {
      required: true,
      query: { searchable: true, filterable: true, sortable: true },
    }),
    prop("currency", stringEnum(["EUR", "USD", "GBP"]), {
      query: { searchable: true, filterable: true, exact: true, facet: true },
    }),
    prop("status", stringEnum(["draft", "sent", "paid", "overdue", "cancelled"]), {
      query: { searchable: true, filterable: true, exact: true, facet: true },
    }),
    prop("issuedAt", "timestamp", {
      query: { searchable: true, filterable: true, sortable: true },
    }),
    prop("dueDate", "date", {
      query: { searchable: true, filterable: true, sortable: true },
    }),
  ],
  search: {
    title: "number",
    exact: ["id", "number"],
  },
  links: [
    link("customer", Customer, { cardinality: "one" }),
    link("project", Project, { cardinality: "one" }),
  ],
})

Flag	Type	Enables
`searchable`	`boolean`	Required gate. Must be `true` before any other flag (or `weight`) applies.
`filterable`	`boolean`	`where(...)` predicates: `eq`, `neq`, ranges, `in`, `exists`, `contains`.
`sortable`	`boolean`	`orderBy(...)` on the property.
`text`	`boolean`	Keyword search over the property via `search(...)`. String-like schemas only.
`exact`	`boolean`	Exact-match search profiles such as `search.exact`.
`facet`	`boolean`	`facets(...)` bucket counts. Field must also be exact-matchable.
`vector`	`boolean`	Vector search on numeric-array embedding fields, when the provider supports it.
`weight`	`number`	Positive relative weight for text ranking. Only valid with `text: true`.

Which schemas support which flag

Each flag is checked against the property's schema at validation time:

Flag	Allowed schemas
`filterable`	exact-matchable schemas, plus `array` and `map`
`sortable`	`string`, `uuid`, `integer`, `double`, `decimal`, `date`, `timestamp`, enums
`text`	`string` and string enums
`exact`	any primitive schema except `fileRef` (so not `object`/`array`/`map`), plus enums
`facet`	same as `exact`
`vector`	numeric arrays (`integer`, `double`, or `decimal` items)

Predicate values are still type-checked against the property schema when the query runs. See Properties for the full schema list.

Object-Type Search Profile

The search config picks which fields a global or type-scoped search uses.

search: {
  title: "company",
  defaultText: ["company", "name", "industry"],
  exact: ["id", "email", "company"],
}

Field	Type	Purpose
`title`	`string`	Display/title property shown in search results. Must be string-like.
`defaultText`	`string[]`	Default keyword-search fields when `search("...")` is called without `fields`.
`exact`	`string[]`	Exact-match fields such as external ids, emails, or invoice numbers.
`vector`	`{ property, source }`	Vector search: `property` stores the embedding, `source` lists the text fields used to produce it.

Every field a profile references must carry the matching property flag:

defaultText fields need text: true
exact fields need exact: true (the primary id is always exact-matchable, so it's exempt)
vector.property needs vector: true, and each vector.source field needs text: true

Search profiles can only reference static properties — telemetry properties (such as Project.progress) are not object-query indexed and will fail validation here.

For vector search, add an embedding property (a numeric array carrying query.vector: true) and point search.vector at it. The source fields must each carry text: true.

search: {
  title: "name",
  defaultText: ["name", "description"],
  vector: { property: "embedding", source: ["name", "description"] },
}

How Metadata Drives Queries

Query call	Required metadata
`where((o) => o.p.status.eq(...))`	`status`: `searchable` + `filterable`
`orderBy(Invoice.p.amount, "desc")`	`amount`: `searchable` + `sortable`
`search("acme")`	`search.defaultText` fields with `searchable` + `text`
`search("acme", { fields: [Customer.p.company] })`	`company`: `searchable` + `text`
`facets([{ property: Invoice.p.status, limit }])`	`status`: `searchable` + `facet` (exact-matchable)
vector search	embedding field: `searchable` + `vector`, plus `search.vector.property`

Missing or mismatched metadata is caught when you call validate() — it reports unknown properties, wrong value types, missing query flags, invalid text fields, and unsupported traversals before any query runs. Provider capability gaps (vector search, relevance sorting) surface at execution time.

Querying Objects — the fluent query builder and predicates.
Properties — property schemas and options.
Object Types — defining object types.
Client Typed Queries — running these queries from the browser.

Search & Query Metadata

Property Query Metadata

Which schemas support which flag

Object-Type Search Profile

How Metadata Drives Queries

Related

Search docs