Use AI Vector Search to Build GenAI Apps with Enterprise Data
summary
Oracle Database 23ai adds built-in AI Vector Search so you can generate embeddings with ONNX models via the VECTOR_EMBEDDING SQL function, store them in a new VECTOR column type, and run similarity queries with either an INMEMORY NEIGHBOR GRAPH (HNSW) for latency-sensitive workloads or NEIGHBOR PARTITIONS for scale-out. It supports target accuracy tuning, an approximate keyword to prefer index scans, transactional consistency, RAC/Exadata scale, RAG workflows with Cohere embeddings in the demo, and a movie-search demo shows vector search outperforms text-theme search.
AI-written summary · about
chapters (15)
- 0:00AI Vector Search intro
- 1:09Vectors capture semantic meaning
- 3:41Converged vs vector databases
- 5:27Retrieval-augmented generation with LLMs
- 6:53Vector storage in Oracle
- 9:44Measuring vector similarity
- 11:31Exact vs approximate search
- 12:56Vector indexes explained
- 15:25Speed vs accuracy
- 17:52Joins and data normalization
- 19:39Multi-vector queries
- 21:26Transactional consistency
- 22:47GenAI pipelines with Oracle
- 23:30Live demo: movie search
- 25:35Summary and advantages
related talks
14mBring AI to Your Data
33mBringing Vector Search to OLTP
57mEnter The Brave New World of GenAI with Vector Search
33mEnter the Brave New World of GenAI with Vector Search
24mFuture of Data and AI
48mEnter the Brave New World of GenAI with Vector Search
01h 21mKeynote on the "AI for Data" Revolution
27mMachine Learning for Advanced AI Search