BigQuery Vector Search: Enabling Semantic Search & RAG LLM

Google Cloud announce today that BigQuery vector search is now generally available (GA), allowing users to search for vector similarity on BigQuery data

Vector indexes are automatically updated using the current k-means model if the underlying data changes

10 billion embeddings can now be indexed, opening the door to large-scale applications

To save money on costly joins when obtaining more data for the search result, you may now save frequently used columns in the index

By changing the base table statement into a query with filters, vector search results can be pre-filtered in conjunction with stored columns

Palo Alto Networks and other clients have used BigQuery vector search to uncover comparable, frequently asked queries, which has sped up the time to insight

BigQuery vector search is becoming an essential part of a multi-modal Retrieval Augmentation Generation (RAG) solution, built on top of a fully functional BigQuery knowledge