Agent-Powered Retrieval with Haystack and Qdrant - Promises and Pitfalls
Qdrant - Vector Database & Search Engine via YouTube
Power BI Fundamentals - Create visualizations and dashboards from scratch
AI, Data Science & Cloud Certificates from Google, IBM & Meta
Overview
Build a Learning Habit
Download Class Central's free printable study calendar
Download for Free
Explore the debate between traditional retrieval pipelines and agent-powered approaches in this 20-minute conference talk from Qdrant's Vector Space Day 2025. Compare conventional retrieval methods with advanced agent-based systems built using Haystack and Qdrant vector database. Learn to construct a traceable, debuggable agent that strategically plans tool usage, selects optimal retrieval strategies, and provides transparent explanations of its decision-making process. Examine practical performance metrics including latency, accuracy, and robustness through hands-on evaluation against a strong non-agent baseline using a movie dataset stored in Qdrant. Master the art of reading agent traces and develop critical decision-making skills to determine when agent-powered retrieval adds genuine value—particularly for complex multi-step tasks, ambiguous queries, or dynamic toolsets—versus when simpler pipeline approaches prove more effective and efficient.
Syllabus
Agent-Powered Retrieval with Haystack and Qdrant: Promises and Pitfalls | deepset | Bilge Yücel
Taught by
Qdrant - Vector Database & Search Engine