Is VectLite an alternative to Pinecone, Weaviate or Chroma?

Yes. VectLite is a single-file embedded vector database (SQLite-style) written in Rust. Unlike Pinecone, Weaviate or Qdrant — which are distributed servers — VectLite runs in-process inside your Python or Node.js app, perfect for local-first and edge use cases.

Does VectLite support hybrid search?

Yes. VectLite combines a dense HNSW index and a BM25 sparse index for hybrid search, with MongoDB-style metadata filters.

What's the difference between VectLite and FAISS?

FAISS is a pure indexing library — no persistence, no transactions. VectLite adds crash-safe persistence, ACID transactions, MongoDB filters and ready-to-use Python/Node.js bindings in a single install.

v0.1.11 — Open Source

vectlite

Embedded vector store for local-first AI

Single-file vector database written in Rust. Dense + sparse hybrid search, HNSW indexing, MongoDB-style metadata filters, transactions and crash-safe persistence — all in a portable .vdb file. No server, no Docker, no network calls.

pip install vectlite

npm install vectlite

Docs PyPI npm GitHub

Features

Hybrid Search

Dense vectors with HNSW + sparse BM25 keyword retrieval, fused with linear combination or RRF.

Single-File Storage

Everything in one portable .vdb file. Crash-safe WAL, file locking, snapshots and backup.

Bulk Ingestion

Import large datasets efficiently with deferred index rebuilds via bulk_ingest() / bulkIngest().

Rich Metadata Filters

MongoDB-style operators: $eq, $gt, $in, $contains, $exists, with nested dot-path access.

Transactions

Atomic batched writes with rollback. Context manager support in Python, try/catch in Node.js.

Multi-Language

Native Rust core with Python (PyO3) and Node.js (napi-rs) bindings.

Quick Start

import vectlite

with vectlite.open("knowledge.vdb", dimension=384) as db:

    db.upsert("doc1", embedding, {"source": "blog", "title": "Auth Guide"})
    db.upsert("doc2", embedding2, {"source": "notes", "title": "Billing"})

    results = db.search(query_embedding, k=5, filter={"source": "blog"})

    # For large imports, prefer bulk_ingest()
    records = [
        {"id": f"doc{i}", "vector": embeddings[i], "metadata": {"source": "corpus"}}
        for i in range(len(embeddings))
    ]
    db.bulk_ingest(records, batch_size=5000)

    for r in results:
        print(r["id"], r["score"])

    db.compact()

Python 3.9+ / Node.js 18+ — Pre-built wheels for macOS, Linux and Windows.

Search & Retrieval

Named Vectors

Multiple vector spaces per record ("title", "body", ...).

Multi-Vector Queries

Weighted search across vector spaces in a single call.

MMR Diversification

Tunable relevance vs. diversity trade-off.

Namespaces

Logical isolation with per-namespace or cross-namespace search.

Built-in Rerankers

text_match(), metadata_boost(), cross_encoder(), bi_encoder() — composable with compose().

Search Diagnostics

search_with_stats() returns timings, BM25 term scores, ANN stats and explain mode.

Data Management

Physical Collections

open_store() manages a directory of independent databases.

Snapshots & Backup

snapshot() creates a self-contained copy. backup() / restore() with ANN sidecars.

Read-Only Mode

Shared locks for safe concurrent readers.

Text Analyzers

Configurable tokenizer pipeline with stopwords (en/fr), Snowball stemming, n-grams.

Language Roadmap

Python — pip install vectlite

Node.js — npm install vectlite

Swift — planned after FFI layer stabilizes

Kotlin — planned after FFI layer stabilizes

MIT License — Maintained by mcsEdition

rustvector-databasehnswbm25hybrid-searchraglocal-firstpythonnodejs

Python 3.9+ · Node.js 18+ · Rust (core)