Skip to content
pubscout

local_hnsw

1.0.0, released 2 years ago by an unverified publisher

A lightweight, in-memory HNSW vector index for Dart and Flutter supporting fast ANN search.

WASM-readyUnverified publisher
$ flutter pub add local_hnsw
Source repository

What it does

Provides an in-memory vector index for performing fast approximate nearest neighbor searches using cosine or Euclidean similarity. Allows adding, searching, deleting, saving, and loading vectors natively in Dart.

AI-written summary of version 1.0.0, based on the package's README and metadata. Report an inaccuracy

Use it when

  • You need on-device vector search or embeddings matching in Dart or Flutter
  • You want a pure Dart implementation without native dependencies
  • You need to save and load small vector indices locally

Skip it when

  • You require a persistent embedded database for massive enterprise-scale vector collections
  • You need complex vector filtering that goes beyond basic distance metrics

Strengths

  • Pure Dart implementation with no native dependencies
  • Supports both cosine and Euclidean similarity metrics
  • Includes built-in save and load serialization methods

Weaknesses

  • In-memory storage means indices must be serialized and reloaded for persistence
  • Not documented for scaling to millions of high-dimensional vectors compared to specialized native vector databases

Setup

Kind
Pure Dart
Effort
Low
Code generation
No
Direct dependencies
0

No native setup or dependencies required.

final index = LocalHNSW<String>(
  dim: 3,
  metric: LocalHnswMetric.cosine,
);
index.add(LocalHnswItem(item: 'apple', vector: [0.1, 0.2, 0.3]));
final result = index.search([0.1, 0.2, 0.3], 2);