dart_vector_index 1.0.1

Fast, pure-Dart vector search & HNSW index for on-device RAG, semantic search, and embeddings with Cosine, Euclidean, SQ8 quantization, and metadata filtering.

1.0.1

• Batch Search Support: Added VectorStore.batchSearch() for high-throughput batch nearest-neighbor queries across multiple embeddings simultaneously.
• Enhanced Test Suite: Added unit test coverage for batch search execution and recall verification.

1.0.0

• Initial stable release of dart_vector_index.
• Pure Dart vector indexing and nearest-neighbor search engine.
• Support for FlatIndex (exact search) and HnswIndex (hierarchical navigable small world graph).
• Multiple distance metrics: cosine, euclidean, dotProduct, manhattan, chebyshev.
• Scalar Quantization (SQ8) for 4x memory reduction on mobile devices.
• Metadata storage and rich predicate filtering (eq, neq, gt, gte, lt, lte, inList, contains, and, or, not).
• High-level VectorStore with atomic JSON serialization, deserialization, batch upsert, and deletion.
• SIMD and Float32List optimized distance math.