waffle_db 0.0.2

An advanced, local vector database for Flutter, powered by Rust and HNSW graph indexing.

0.0.2

• HNSW Save Reliability: Fixed a bug where closing or flushing an empty HNSW index returned an error by skipping index dump if the graph contains 0 elements.

• Index Load Resiliency: Fixed a startup crash ("failed to fill whole buffer") by ensuring both the map and the HNSW graph file exist on disk before trying to load them.

• Asynchronous Benchmark Suite: Added a comprehensive real-world async performance benchmark suite under test/vector_db_real_world_benchmark_test.dart using benchmark_harness.

• Color Semantic Search Example: Rewrote the example app under example/lib/main.dart into an interactive, visual color semantic search demonstration that inserts up to 12 million random colors and queries the 150 closest matches in milliseconds.

• Performance Optimization: Optimized vector serialization/deserialization on read paths using fast raw pointer copies.

• Memory Footprint Reduction: Pre-allocated internal map capacities (id_map, reverse_id_map, and ID list vectors) based on database size, preventing multiple heap reallocations.

• Mutex Lock Consolidation & Deadlock Prevention: Combined separate ID mapping locks into a single IdRegistry mutex, avoiding cross-lock deadlock hazards and minimizing locking overhead.

• Zero-Copy Batch Insertion: Refactored waffle_insert_batch to slice the flat array directly for disk persistence, reducing heap-allocated vector clones by 50%.

• Correctness & Performance Fix: Linked includeMetadata in WaffleQueryBuilder to the underlying Rust API so metadata is correctly retrieved when requested, and omitted when not needed.

• Dart Namespace Caching: Cached collection namespace prefixes in WaffleCollection to avoid dynamic string creation during filtering.

0.0.1+1

• fix unsupported platforms

0.0.1

• Initial release of Waffle-DB.