"Qdrant Vector Search: Shipping High-Recall Retrieval with Filters and Payloads"
Modern retrieval systems fail in subtle ways: they look fast in demos, then lose critical results once filters, payloads, tenancy rules, and production latency targets arrive. This book is for experienced search, ML, and backend engineers who need more than a vector database introduction. It treats Qdrant as a retrieval engine, showing how architecture, schema, indexing, and operations combine to determine whether real systems actually return the right candidates under pressure.
Across the book, readers learn how to model collections, points, vectors, and payloads for retrieval-first design; how vector similarity, ANN behavior, and filter selectivity interact; and how to tune both vector and payload indexes for high recall. It also covers the unified Query API, hybrid and multi-stage retrieval patterns, rigorous evaluation for filtered workloads, and the operational realities of running Qdrant in production. The emphasis is on measurable quality, practical trade-offs, and decision-making under constraints.
Rather than repeating basic concepts, the book follows a deliberate progression from system model to tuning, evaluation, and scale. Readers should already be comfortable with embeddings, search fundamentals, and production software systems. In return, they get a focused, deeply technical guide to building retrieval systems that remain accurate, explainable, and resilient when real-world metadata and operational complexity enter the loop.
