LibSQL/Turso storage and hybrid search backend for Worlds.

Standalone LibSQL package extracted from @worlds/sdk.

Install

Package managers

# Deno (first-class JSR support)
deno add jsr:@worlds/libsql

# Bun / npm / pnpm / Yarn (via JSR npm compatibility layer)
npx jsr add @worlds/libsql

CDN (browser / no build step)

esm.sh serves JSR packages as ES modules — no install, no bundler needed.

import { createLibsqlWorldsSdk } from "https://esm.sh/jsr/@worlds/libsql@0.4.1";

With an import map:

<script type="importmap">
{
  "imports": {
    "@worlds/libsql": "https://esm.sh/jsr/@worlds/libsql@0.4.1"
  }
}
</script>

Then a module script:

import { createLibsqlWorldsSdk } from "@worlds/libsql";

Pin to an exact build for deterministic caching:

import { createLibsqlWorldsSdk } from "https://esm.sh/jsr/@worlds/libsql@0.4.1?pin=v1724100000";

Usage

Full client: quad store + search index + SPARQL

import { createClient } from "@libsql/client";
import { createLibsqlWorldsSdk } from "@worlds/libsql";

const databaseClient = createClient({ url: ":memory:" });
// The factory assembles the three strategy objects internally: a
// LibsqlConnectionDriver over the raw client, the schema builder, and the
// search-query builder. Callers just pass the LibSQL client.
const client = await createLibsqlWorldsSdk({ client: databaseClient });

await client.import({
  source: {
    kind: "serialized",
    data:
      `<http://example.com/alice> <http://example.com/knows> <http://example.com/bob> .`,
    contentType: "text/turtle",
  },
});

const { search, sparql } = await Promise.all([
  client.search({ query: "alice" }),
  client.sparql({
    query:
      `SELECT ?o WHERE { <http://example.com/alice> <http://example.com/knows> ?o }`,
  }),
]);

SPARQL only: the RDF/JS store + WazooSparqlEngine over LibSQL

When you only need SPARQL over an RDF/JS store backed by LibSQL (no chunk search, no embeddings), wire LibsqlRdfjsStore directly into @wazoo/sparql-engine's WazooSparqlEngine:

import { createClient } from "@libsql/client";
import type * as rdfjs from "@rdfjs/types";
import { WazooSparqlEngine } from "@wazoo/sparql-engine";
import {
  initializeLibsqlSchema,
  LibsqlConnectionDriver,
  LibsqlQuadStore,
  LibsqlRdfjsStore,
  LibsqlSchemaBuilder,
  LibsqlSearchQueryBuilder,
} from "@worlds/libsql";

// A raw LibSQL client — in-memory here, any remote/embedded URL works.
const databaseClient = createClient({ url: ":memory:" });

// Create the quads table + covering indexes (no FTS/vector chunk schema).
const schemaBuilder = new LibsqlSchemaBuilder(32);
const connection = new LibsqlConnectionDriver(databaseClient);
await initializeLibsqlSchema(connection, schemaBuilder);

// The RDF/JS read source over LibSQL: match/countQuads via SQL index seeks.
const store = new LibsqlRdfjsStore({ connection });

// A quad store for writes (import/export/transaction); chunk projection
// is skipped by leaving searchIndexProjector unset.
const quadStore = new LibsqlQuadStore({
  connection,
  store,
  searchQueryBuilder: new LibsqlSearchQueryBuilder(32),
});

// SPARQL engine over the same LibSQL-backed store.
const sparqlEngine = new WazooSparqlEngine({
  store: store as unknown as rdfjs.Store,
  createTransaction: () => quadStore.createTransaction(),
});

await quadStore.import({
  source: {
    kind: "serialized",
    data:
      `<http://example.com/alice> <http://example.com/knows> <http://example.com/bob> .`,
    contentType: "text/turtle",
  },
});

const result = await sparqlEngine.execute({
  query:
    `SELECT ?o WHERE { <http://example.com/alice> <http://example.com/knows> ?o }`,
});

Development

This package consumes the shared, driver-free SQL plan layer from @worlds/sqlite/sql-core — the FTS5 sanitizer/stopwords, buildChunkFtsValue, buildSearchResultId, the chunks_fts DDL/triggers, and filter-clause helpers. The LibSQL-specific dialect (native vector32 columns, in-SQL RRF fusion over vector_top_k('idx_chunks_vector', ...), and the column-per-position quads table) stays local to this repo.

deno task ci

Dry-run a JSR publish locally:

deno task publish:dry

Publishing to JSR

Releases publish automatically when changes merge to main. Bump "version" in deno.json in each release PR — JSR rejects duplicate versions.

One-time setup on jsr.io/@worlds/libsql:

1. Open package settings and link https://github.com/wazootech/worlds-libsql.
2. Enable GitHub Actions publishing (OIDC). The publish workflow uses id-token: write; no JSR token secret is required when OIDC is configured.
3. Confirm your GitHub account can publish to the @worlds org.

After setup, merging to main runs CI, a publish dry-run, and deno publish.