Uni: Embedded Graph & Vector Database

Uni is a modern, embedded database that combines property graph (OpenCypher), vector search, and columnar storage (Lance) into a single, cohesive engine. It is designed for applications requiring local, fast, and multimodal data access, backed by object storage (S3/GCS) durability.

Part of The Rustic Initiative by Dragonscale Industries Inc.

Key Features

• Embedded & Serverless: Runs as a single process or library within your application.
• Multimodal: Supports Graph, Vector, Document, and Columnar workloads.
• Storage: Persists data using LanceDB format on local disk or object storage.
• Query Language: OpenCypher (subset) with vector search extensions.
• Search: Vector similarity, full-text (BM25), and hybrid search with rank fusion.
• Algorithms: Built-in graph algorithms (PageRank, Louvain, ShortestPath, etc.).
• Connectivity: Rust crate and Python bindings.

Getting Started

Rust

Add uni-db to your Cargo.toml:

[dependencies]
uni-db = { git = "https://github.com/rustic-ai/uni" }

use uni_db::Uni;

#[tokio::main]
async fn main() -> anyhow::Result<()> {
    let db = Uni::open("my_db").build().await?;
    
    // Create node
    db.execute("CREATE (n:Person {name: 'Alice', age: 30})").await?;
    
    // Query
    let results = db.query("MATCH (n:Person) RETURN n").await?;
    println!("{:?}", results);
    
    Ok(())
}

Python

Build and install the python bindings:

cd bindings/uni-db
maturin develop

import uni_db

db = uni_db.Database("my_db")

# Create
db.query("CREATE (n:Person {name: 'Bob', age: 25})")

# Query
results = db.query("MATCH (n:Person) RETURN n")
print(results)

Search Capabilities

Uni provides three powerful search methods that can be combined with graph traversals:

Vector Search

-- Search by semantic similarity (auto-embeds text if embedding config exists)
CALL uni.vector.query('Document', 'embedding', 'machine learning tutorial', 10)
YIELD node, score
RETURN node.title, score

Full-Text Search (BM25)

-- Traditional keyword search with relevance scoring
CALL uni.fts.query('Article', 'content', 'database optimization', 20)
YIELD node, score
RETURN node.title, score

Hybrid Search (Vector + FTS Fusion)

-- Combine semantic and keyword search with Reciprocal Rank Fusion
CALL uni.search(
    'Document',
    {vector: 'embedding', fts: 'content'},
    'neural network architectures',
    null,  -- auto-embed the query
    10
)
YIELD node, score, vector_score, fts_score
RETURN node.title, score

See Language Extensions for full procedure documentation.

Documentation

• Full Documentation
• Architecture Design
• Feature Comparison
• Algorithms

Contributing

We welcome contributions! Please see our Contributing Guide for details.

License

Apache 2.0 - see LICENSE for details.

---

Uni is developed by Dragonscale Industries Inc. as part of The Rustic Initiative.