Overview
BVector is BangDB's purpose-built vector database engine — fast ANN search, flexible indexing, and a clean REST API for AI-native applications.
What is BVector?
BVector is the vector storage and retrieval layer of the BangDB platform. It handles ingestion, indexing, and querying of high-dimensional embedding vectors — the core data type behind semantic search, recommendation engines, and RAG (Retrieval Augmented Generation) pipelines.
Unlike general-purpose databases bolted with vector extensions, BVector is designed from the ground up for embedding workloads. It uses HNSW indexing and supports cosine, Euclidean, and dot product similarity metrics out of the box.
Key capabilities
ANN Search
Approximate nearest neighbor search with sub-millisecond query latency at any scale.
HNSW Indexing
Hierarchical Navigable Small World graphs — the gold standard for high-recall vector indexing.
On-Premise
Full data sovereignty. Deploy on your own infrastructure, air-gapped if needed.
REST API
Simple HTTP API for index management, vector upserts, and similarity queries.
Multi-Index
Run multiple isolated vector indexes on a single instance with independent configurations.
Real-time Ingestion
Ingest and query vectors simultaneously with no index rebuild required.
How BVector and BVector Chat relate
BVector Chat is a pre-built UI layer on top of BVector. When you embed the chat widget, it connects to your BVector instance to perform semantic retrieval and generate context-aware responses. BVector is the engine; BVector Chat is the interface.