May 2024 Summaries

The author recently organized meetups for Supabase's GA Week at over 30 locations worldwide, featuring an interactive quiz game built using a simplified version of Kahoot, with features such as anonymous sign-in authentication through Supabase's auth feature, database storage for questions and answers, and real-time updates to keep the game state in sync. The app is built using Next.js and utilizes three core Supabase features: auth, database, and realtime. The quiz game has two screens, a host screen and a player screen, with the host starting the game by choosing a quiz set and players answering questions on their phones to earn points. The leaderboard is displayed on the screen once all questions are answered, and the app can be found in the Supabase GitHub repository, where contributors are welcome to receive pull requests and contribute to the open-source project.

The latest version of pgvector introduces several approaches to leverage redundancy in real-world embedding datasets, leading to memory and performance savings with minimal impact on precision. These include float16 vector representation, sparse vectors, bit vectors, and new distance functions such as L1, Hamming, and Jaccard distances. The use of half vectors (float16) can reduce memory consumption by 50%, while sparse vectors save significant storage space for vectors with many zero components. Bit vectors allow for fast pre-selection from a dataset before performing an additional search within the subset. These improvements have led to over 100x speedup compared to one year ago, making pgvector more efficient and effective in handling real-world embedding datasets.