The Hitchhiker's Guide to Vector Search - Qdrant

This post provides practical advice and real-world examples for optimizing textual RAG pipelines. It details strategies for text extraction (object-based vs. agentic/OCR), chunking (sentence/token-based, semantic, agentic/neural), and embedding choices (dense vs. sparse, hybrid search with rescoring/fusion). It also covers search boosting techniques like semantic caching and binary quantization, and discusses query optimization through transformation (expansion, decomposition) based on query type. The author shares personal project examples to illustrate these concepts.