TL;DR — Vector Database

Design a vector database from scratch that stores high-dimensional embeddings and supports fast approximate nearest neighbor (ANN) search. The system must handle billions of vectors across distributed nodes while maintaining sub-50ms query latency. Key features: Insert, update, and delete vectors with metadata. K-nearest neighbor search with cosine, euclidean, or dot product distance.

HARD60 min

Vector Database

HNSWANNvector searchshardingindexingsimilarity

Key Points

Insert, update, and delete vectors with metadata
K-nearest neighbor search with cosine, euclidean, or dot product distance
Metadata filtering combined with vector similarity (pre-filter or post-filter)

Key Constraints

Vector dimensions

768-1536

Total vectors

1B+

Index size (1B x 1536d)

~6TB in memory

Hints (0/3)

Canvas

Build your design

Drag components from the palette to build your solution for "Vector Database"

Vector Database

▶3D Simulate

Advanced60 min read+200 XP

TL;DR — Vector Database

HARD60 min

Vector Database

HNSWANNvector searchshardingindexingsimilarity

Key Points

Insert, update, and delete vectors with metadata
K-nearest neighbor search with cosine, euclidean, or dot product distance
Metadata filtering combined with vector similarity (pre-filter or post-filter)

Key Constraints

Vector dimensions

768-1536

Total vectors

1B+

Index size (1B x 1536d)

~6TB in memory

Hints (0/3)

Canvas

Build your design

Drag components from the palette to build your solution for "Vector Database"