Search

QUESTPIE supports full-text search using PostgreSQL's built-in FTS capabilities and optional embedding-based semantic search with pgvector.

Currently, PostgreSQL FTS and pgvector are the only implemented search adapters. External services like Meilisearch or Elasticsearch are not yet supported.

Making Collections Searchable

Add .searchable() to index a collection for full-text search:

src/questpie/server/collections/posts.ts

import { collection } from "#questpie/factories"

export default collection("posts")
  .fields(({ f }) => ({
    title: f.text(255).required(),
    body: f.textarea(),
    tags: f.text(500),
  }))
  .searchable({
    content: (record) => `${record.title} ${record.body ?? ""} ${record.tags ?? ""}`,
    metadata: (record) => ({
      published: record.status === "published",
    }),
  })

Option	Type	Description
`content`	`(record) => string`	Extracts plain text for full-text indexing. Called on create/update.
`metadata`	`(record) => object`	Structured metadata for faceted filtering.

How It Works

When a record is created or updated:

The content function extracts plain text from the record
PostgreSQL generates a tsvector for the text using the configured language
A GIN index enables fast full-text queries
BM25 ranking orders results by relevance

Client Search

const results = await client.search.search({
  query: "haircut styles",
  collections: ["posts", "services"],
  limit: 20,
})

Search Options

Option	Type	Description
`query`	`string`	Search query (supports PostgreSQL tsquery syntax)
`collections`	`string[]`	Limit search to specific collections
`limit`	`number`	Maximum results (default: 50)

Server-Side Search

From routes or hooks:

handler: async ({ search }) => {
  const results = await search.search({
    query: "booking confirmation",
    collections: ["posts"],
  })
  return results
}

Embedding-Based Search (pgvector)

For semantic search using vector embeddings:

.searchable({
  content: (record) => record.title + " " + record.body,
  embeddings: {
    enabled: true,
    dimensions: 1536,
    model: "text-embedding-3-small",
  },
})

This stores vector embeddings alongside the text index, enabling similarity-based queries that understand meaning rather than just keywords.

Collections — Data modeling
Querying — Query operators