HollowDB Vector

Overview

HollowDB Vector is a decentralized vector database for building vector search applications, developed by FirstBatch Labs. It implements Hierarchical Navigable Small Worlds (HNSW) index over HollowDB, combining decentralized storage with vector search capabilities.

Core Technology

Built on HollowDB, a decentralized privacy-preserving key-value database on Arweave network, powered by Warp Contracts. The system provides a database for applications with production-level stability and feasibility for projects of any scale, while also being privacy-preserving.

Key Features

• Decentralized Architecture: Runs on Arweave's permaweb for permanent, censorship-resistant storage
• Privacy-Preserving Operations: Data update and remove operations gated with zero-knowledge proofs
• HNSW Indexing: Implements state-of-the-art HNSW algorithm for efficient vector similarity search
• Performance: Write interaction ~1.5s on average, read interaction ~300ms on average with gRPC (bundled data)
• Zero-Knowledge Proofs: Users can only update key-value pairs by providing a zero-knowledge proof

Use Cases

• Web3 AI applications requiring decentralized vector search
• Privacy-sensitive applications needing permanent, verifiable storage
• Decentralized knowledge bases and semantic search engines
• AI applications that need censorship-resistant vector databases

Licensing

HollowDB Vector is licensed under Apache 2.0, making it free and open-source for commercial and non-commercial use.

Pricing

Open-source and free to use. Storage costs are based on Arweave network fees for permanent data storage.