NGT-labs/NGT

Health v2 · maintenance only

Adoption not part of the score

1374 stars · 131 forks observed · 2026-10-02

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

NGT is a C++ library and command-line tool from Yahoo Japan for high-speed approximate nearest neighbor search on high-dimensional vector data, using neighborhood graph and tree-based indexing with quantized variants (QG, QBG). It offers a Python package via PyPI and supports distance metrics like L2, cosine, Jaccard, and inner product.

Use cases

• find similar vectors in a high-dimensional dataset
• build an approximate nearest neighbor index for embeddings
• search k-nearest neighbors for image or text embeddings
• speed up similarity search over millions of vectors
• use quantized graph indexes to reduce memory for vector search
• integrate ANN search into a Python application

When to choose

• you need fast approximate nearest neighbor search on tens to thousands of dimensions
• you want a mature C++ core with Python bindings
• memory-efficient quantized indexes (QG/QBG) are needed for large datasets

When to avoid

• you need a full managed vector database with filtering and replication
• your data is small enough that brute-force search suffices
• you need first-class Windows support or non-Linux/macOS deployment

Facets

library · maturity active

vector-database search-engine machine-learning cpp python cli approximate-nearest-neighbor-search ann vector-search graph-index quantization high-dimensional-data search algorithms linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="NGT-labs/NGT")