OmenDB
v0.0.11

Vector search without the infrastructure

Embedded vector database for Python and Node.js. No server. No Docker. No configuration. Just pip install.

$ pip install omendb
Documentation
rag_example.py 
import omendb

# Create database — no server, no setup
db = omendb.open("./vectors", dimensions=768)

# Add vectors with metadata
db.set([{
    "id": "doc_1",
    "vector": embedding,  # from any embedding model
    "text": "Your document text here",
    "metadata": {"category": "docs"}
}])

# Search with optional filtering
results = db.search(query_embedding, k=10)
results = db.search(query_embedding, k=10, filter={"category": "docs"})

Why embedded?

Skip the infrastructure. Your data stays local, latency stays low, and cloud bills stay at zero.

Data stays local

Your vectors never leave your machine. Perfect for sensitive data, air-gapped environments, or when you just don't want another vendor.

Sub-millisecond latency

No network round-trips. Search runs in-process with your application. HNSW + SIMD gives you 10,000+ queries per second.

$0 cloud bills

No managed service fees. No per-query pricing. No usage tiers. Just a library that runs wherever Python or Node.js runs.

Performance

Search queries per second with 10K vectors

128 dimensions
10K+
queries/sec
768 dimensions
3K+
queries/sec
1536 dimensions
1.5K+
queries/sec

Tested on Apple M3 Max and Intel i9-13900KF · Batch queries achieve 6-8x higher throughput

Features

Vector Search

Find similar vectors fast using HNSW. Supports L2, cosine, and dot product distance. Tune speed vs accuracy to your needs.

Hybrid Search

Combine keyword matching with semantic similarity. Get the best of traditional search and vector search in one query.

Metadata Filtering

Filter results by any field in your metadata. Combine conditions with AND/OR. Stays fast even with selective filters.

Quantization

Reduce memory 4-8x while keeping 98-99% accuracy. One flag to enable, automatic rescoring keeps results precise.

Persistence

Data saved to a single file, crash-safe by default. Fast startup even with millions of vectors. Or use in-memory for testing.

Multi-language

Native support for Python, Node.js, and Rust. Same simple API across all languages.

Works with your stack

🦜 LangChain
🦙 LlamaIndex
OpenAI
</> Any embedding model

Get started in seconds

Python
pip install omendb
Node.js
npm install omendb
Rust
cargo add omendb