OmenDB

Embedded vector database for Python and Node.js.
No server. No setup. Just install.

$ pip install omendb

example.py

import omendb

# Open or create database
db = omendb.open("./vectors", dimensions=768)

# Add vectors with metadata
db.set([{
    "id": "doc1",
    "vector": embedding,
    "text": "Hello world",
    "metadata": {"category": "greeting"}
}])

# Search with filters
results = db.search(query_vector, k=10, filter={"category": "greeting"})

No Server

Runs in-process. No Docker, no ports, no configuration.

Persistent

Data saved automatically. Survives restarts.

Fast

HNSW index with SIMD. 10K+ QPS on laptop.

Batteries Included

Hybrid search, filtering, quantization built-in.

Features

Vector Search

HNSW graph with 95%+ recall. L2, cosine, and dot product distance.

Hybrid Search

BM25 text search combined with vector similarity using RRF fusion.

Filtering

MongoDB-style metadata queries with ACORN-1 algorithm.

Quantization

4-8x compression with ~99% recall. Reduce memory usage.

Collections

Multiple namespaces in a single database file.

Multi-language

Python, Node.js, and Rust. Same API everywhere.

Installation

Python

pip install omendb

Node.js

npm install omendb

Rust

cargo add omendb