Skip to main content

libSQL

Turso is a SQLite-compatible database built on libSQL, the Open Contribution fork of SQLite. Vector Similiarity Search is built into Turso and libSQL as a native datatype, enabling you to store and query vectors directly in the database.

LangChain.js supports using a local libSQL, or remote Turso database as a vector store, and provides a simple API to interact with it.

This guide provides a quick overview for getting started with libSQL vector stores. For detailed documentation of all libSQL features and configurations head to the API reference.

Overview​

Integration details​

ClassPackageJS supportPackage latest
LibSQLVectorStore@langchain/community✅npm version

Setup​

To use libSQL vector stores, you'll need to create a Turso account or set up a local SQLite database, and install the @langchain/community integration package.

This guide will also use OpenAI embeddings, which require you to install the @langchain/openai integration package. You can also use other supported embeddings models if you wish.

You can use local SQLite when working with the libSQL vector store, or use a hosted Turso Database.

tip

See this section for general instructions on installing integration packages.

npm install @libsql/client @langchain/openai @langchain/community

Now it's time to create a database. You can create one locally, or use a hosted Turso database.

Local libSQL​

Create a new local SQLite file and connect to the shell:

sqlite3 file.db

Hosted Turso​

Visit sqlite.new to create a new database, give it a name, and create a database auth token.

Make sure to copy the database auth token, and the database URL, it should look something like:

libsql://[database-name]-[your-username].turso.io

Setup the table and index​

Execute the following SQL command to create a new table or add the embedding column to an existing table.

Make sure to mopdify the following parts of the SQL:

  • TABLE_NAME is the name of the table you want to create.
  • content is used to store the Document.pageContent values.
  • metadata is used to store the Document.metadata object.
  • EMBEDDING_COLUMN is used to store the vector values, use the dimensions size used by the model you plan to use (1536 for OpenAI).
CREATE TABLE IF NOT EXISTS TABLE_NAME (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    content TEXT,
    metadata TEXT,
    EMBEDDING_COLUMN F32_BLOB(1536) -- 1536-dimensional f32 vector for OpenAI
);

Now create an index on the EMBEDDING_COLUMN column:

CREATE INDEX IF NOT EXISTS idx_TABLE_NAME_EMBEDDING_COLUMN ON TABLE_NAME(libsql_vector_idx(EMBEDDING_COLUMN));

Make sure to replace the TABLE_NAME and EMBEDDING_COLUMN with the values you used in the previous step.

Instantiation​

To initialize a new LibSQL vector store, you need to provide the database URL and Auth Token when working remotely, or by passing the filename for a local SQLite.

import { LibSQLVectorStore } from "@langchain/community/vectorstores/libsql";
import { OpenAIEmbeddings } from "@langchain/openai";
import { createClient } from "@libsql/client";

const embeddings = new OpenAIEmbeddings({
  model: "text-embedding-3-small",
});

const libsqlClient = createClient({
  url: "libsql://[database-name]-[your-username].turso.io",
  authToken: "...",
});

// Local instantiation
// const libsqlClient = createClient({
//  url: "file:./dev.db",
// });

const vectorStore = new LibSQLVectorStore(embeddings, {
  db: libsqlClient,
  tableName: "TABLE_NAME",
  embeddingColumn: "EMBEDDING_COLUMN",
  dimensions: 1536,
});

Manage vector store​

Add items to vector store​

import type { Document } from "@langchain/core/documents";

const documents: Document[] = [
  { pageContent: "Hello", metadata: { topic: "greeting" } },
  { pageContent: "Bye bye", metadata: { topic: "greeting" } },
];

await vectorStore.addDocuments(documents);

Delete items from vector store​

await vectorStore.deleteDocuments({ ids: [1, 2] });

Query vector store​

Once you have inserted the documents, you can query the vector store.

Query directly​

Performing a simple similarity search can be done as follows:

const resultOne = await vectorStore.similaritySearch("hola", 1);

for (const doc of similaritySearchResults) {
  console.log(`${doc.pageContent} [${JSON.stringify(doc.metadata, null)}]`);
}

For similarity search with scores:

const similaritySearchWithScoreResults =
  await vectorStore.similaritySearchWithScore("hola", 1);

for (const [doc, score] of similaritySearchWithScoreResults) {
  console.log(
    `${score.toFixed(3)} ${doc.pageContent} [${JSON.stringify(doc.metadata)}`
  );
}

API reference​

For detailed documentation of all LibSQLVectorStore features and configurations head to the API reference.

Was this page helpful?


You can also leave detailed feedback on GitHub.