sqlite-wasm-vec

library0.1.11jsnpmunverified

SQLite compiled to WebAssembly with the sqlite-vec extension for vector similarity search. Version 0.1.11 bundles SQLite 3.51.0 and sqlite-vec v0.1.7-alpha.2 in a ~1.5MB WASM binary. Provides full SQLite API with vector operations (k-NN search, distance functions) directly in browsers and Node.js. Supports float, int8, and bit vector types via virtual tables. Includes OPFS for persistent storage in browsers and first-class TypeScript types. Fork of @sqlite.org/sqlite-wasm with all original features intact.

database ai-ml

npm install sqlite-wasm-vec

github ↗package ↗homepage ↗

INSTALL

IMPORT

SIG · SQLITE-WASM-VEC

sqlite-wasm-vec

databasejavascriptv0.1.11

harness data pending

Install & Compatibility

Where this runs

No compatibility data collected yet for this library.

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

sqlite3InitModule

✓ import sqlite3InitModule from 'sqlite-wasm-vec'

✗ import { sqlite3InitModule } from 'sqlite-wasm-vec'

Default export only. Named import will cause undefined.

oo1.DB

✓ const db = new sqlite3.oo1.DB(':memory:')

✗ import { DB } from 'sqlite-wasm-vec'

DB is a property of the sqlite3 object returned by init; not directly importable.

sqlite3InitModule (CommonJS)

✓ const sqlite3InitModule = require('sqlite-wasm-vec').default

✗ const sqlite3InitModule = require('sqlite-wasm-vec')

CJS users must use .default; ES module is the primary format.

Initializes SQLite with vector support, creates a virtual table, inserts a vector, and performs a k-NN similarity search.

import sqlite3InitModule from 'sqlite-wasm-vec'; const sqlite3 = await sqlite3InitModule(); const db = new sqlite3.oo1.DB(':memory:'); // Create virtual table for vectors db.exec(`CREATE VIRTUAL TABLE vec_items USING vec0(embedding float[384])`); // Insert a vector const embedding = new Float32Array(384).fill(0.1); db.exec({ sql: 'INSERT INTO vec_items(rowid, embedding) VALUES (?, ?)', bind: [1, embedding.buffer] }); // Search for nearest neighbors const query = new Float32Array(384).fill(0.2); const results = db.selectArrays( 'SELECT rowid, distance FROM vec_items WHERE embedding MATCH ? ORDER BY distance LIMIT 5', query.buffer ); console.log('Results:', results);

Debug

Known issues

gotchaNode.js support is limited to in-memory databases. Persistent storage is not available in Node.js.

fix

Use ':memory:' for Node.js; for persistence, run in browser with OPFS enabled.

affects: >=0.1.0

gotchaVector buffers must be passed as ArrayBuffer (typedArray.buffer) to binding functions or exec bind. Passing Float32Array directly will fail.

fix

Use `embedding.buffer` when binding vector data to SQL statements.

affects: >=0.1.0

deprecatedThe default export is the initialization function alone. There is no separate export for the WASM URL or binary.

fix

Access the WASM resource internally; no public export provided.

affects: >=0.1.0

gotchasqlite-vec functions (vec_version, vec_distance_cosine, etc.) only work after initialization and on tables created with `vec0` virtual table syntax.

fix

Ensure you create virtual tables using `CREATE VIRTUAL TABLE ... USING vec0(...)` before calling vector operations.

affects: >=0.1.0

Errors

Common errors & fixes

Error: vec0 virtual table not found

Forgot to include sqlite-vec extension or created regular table instead of virtual table.

fix

Use `CREATE VIRTUAL TABLE ... USING vec0(...)` syntax.

TypeError: Cannot read properties of undefined (reading 'oo1')

Importing named export 'sqlite3' instead of default export from 'sqlite-wasm-vec'.

fix

Use `import sqlite3InitModule from 'sqlite-wasm-vec'` and then `const sqlite3 = await sqlite3InitModule()`.

Error: The argument to bind must be an ArrayBuffer

Passing Float32Array directly instead of its buffer property to SQL bind parameters.

fix

Pass `embedding.buffer` instead of `embedding`.

Upgrade

Version history

0.1.11latest on npm

Audit

Dependencies

No dependency data recorded yet.

Agent activity

23 hits · last 30 days

node