Build your own search application with Weaviate
summary
Overview of an open-source vector database that stores data as high-dimensional embeddings and runs semantic search, question answering and classification using pre-trained encoders (BERT/sentence-transformers for text, ResNet50 for images) and HNSW approximate nearest neighbor indexing. Live demos index a small news corpus (~4,000 articles) and the full English Wikipedia, return results ranked by cosine distance and a certainty score (0–1), support near-text search and multimodal queries, and layer reader/generator models for exact answer extraction and summarization.
AI-written summary · about
chapters (20)
- 0:00Vector search engines intro
- 1:10Unstructured data challenges
- 3:15Traditional search failures
- 4:59Semantics with vector search
- 5:40Google as vector search
- 7:48Need for custom vector search
- 8:50Vectors represent meaning
- 10:57Vector vs graph databases
- 13:47Vector search workflow
- 16:13Storing and retrieving vectors
- 17:59Simple Weaviate search pipeline
- 18:40Extending with reader/generator
- 20:50Weaviate API and GraphQL
- 21:33Live demo: semantic news search
- 25:53Question answering with readers
- 28:04Wikipedia search with Weaviate
- 28:26Multi-modal text and image search
- 29:09ML models in Weaviate
- 29:51Automatic vector classification
- 30:14Closing and resources
related talks
24mIntroduction to Vector Search with Weaviate
13mSearch through your data with Weaviate
38mAI-powered Vector Search Engine Weaviate
23mWeaviate Vector Search Engine – Introduction
27mUsing Weaviate for building Generative AI apps
45mA gentle introduction to Vector Databases
11mLightning Talk: Weaviate Vector Search Engine - Enhancing Cybersecurity
57mUsing Vector Databases for Multimodal Embeddings and Search