Practical Vector Search Configuration, Tuning, RAG, and the Road Ahead
summary
Walks through a working Drupal setup using Search API + Drupal AI with vector stores like Milvus or PGVector (Milvus ≈ Pinecone) and embedding models via Llama Server or the open‑weight embedding Gema. Covers chunking, overlapping, field vs context vs filter placement, and hybrid filtering quirks; indexing a ~10,000‑node dataset takes about 9 hours to index ~10,000 nodes on an M1 MacBook Air, memory limits appear in ddev, and basic RAG uses 10 retrieved items per query plus an LLM for final generation.
AI-written summary · about
chapters (15)
- 0:00Introduction and background
- 1:28What is vector search
- 3:41Getting started with realistic data
- 5:53Setting up vector search in Drupal
- 9:23Indexing and chunking strategies
- 12:16Testing and indexing process
- 15:31Testing search results
- 16:15Filtering and facets limitations
- 19:07Retrieval augmented generation
- 23:20Tuning and relevance testing
- 25:29Performance and scaling challenges
- 27:17Hybrid search opportunities
- 28:43Drupal ecosystem and hosting
- 29:25Key takeaways and future outlook
- 31:12Q&A on costs and caching
related talks
33mVector Databases in the Age of AI: Embracing RAG for Enhanced Search Capabilities
04mSupercharging GenAI PostgreSQL Vector Search Optimization
47mBoosting AI With OpenSearch: Vector Stores, Similarity Search, and Retrieval-Augme...
01h 35mTutorial: Build AI-Supercharged RAG Apps with a Vector Database
36mUnleashing the Power of Vector Databases for AI-Driven Applications
33mEnter the Brave New World of GenAI with Vector Search
28mA Deep Dive into Python-Powered Precision and Scalability
48mEnter the Brave New World of GenAI with Vector Search