Don’t Get Lost in Vector Space
summary
An Italian open-source AI project, Cheshire Cat, uses the Qdrant vector database to implement Retrieval-Augmented Generation (RAG) memories for LLM-powered apps. The talk explains managing three separate memories (conversations, documents, tools), running Qdrant in-memory/file/container for testing/dev/prod, handling vector-dimension changes when switching embedders, and using snapshots for just-in-time backups, plus practical challenges and Qdrant features used to solve them.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:16Vanilla RAG basics
- 5:47Vector space
- 7:12Memory design in Cerchat
- 8:37Drunken Cat Effect
- 10:26Embedder changes with aliases
- 12:11Vector storage and quantization
- 16:50Matryoshka embeddings
- 19:182D vector space mapping
- 21:49Document export and management
- 23:53Q&A
related talks
01h 35mTutorial: Build AI-Supercharged RAG Apps with a Vector Database
31mRetrieval Augmented Generation
13mRAG Without the Hassle: Building AI-Powered Applications
50mRetrieval Augmented Generation Beyond Toy Demos
27mWhen Vectors Aren’t Enough: Unlocking AI’s Hidden Connections with Graphs
48mEnter the Brave New World of GenAI with Vector Search
17mThe keys to success with Generative AI and Vector Databases
48mA Fun & Absurd Introduction to Vector Databases