Entity-graph retrieval on top of Qdrant: Personalized PageRank + convergence scoring, zero LLM calls at query time, +5.9 pts recall@10 on full HotpotQA.
Qdrant answers "which vectors are near this query?" in milliseconds at
billion scale. But there's a class of questions where nearness is the
wrong criterion entirely:
"Where did the founder of the company that acquired Slack study?"
The passage that answers this talks about Marc Benioff and USC. It
Multi-hop retrieval over Chroma with no LLM in the query loop: deterministic, ~100ms, benchmarked on HotpotQA and MuSiQue.
HNSW turns nearest-neighbor search from brute-force (2M comparisons, 1.2s) into a hierarchical graph walk (1.8k comparisons,…
IVF's cluster-and-probe vs HNSW's small-world graph — recall, memory, build time, and how each lands in pgvector and FAISS.
There's a moment in almost every RAG project where someone asks the question that decides your next...
Context graphs traverse entity relationships ANN search can't follow. Learn when to use graphs, vectors, or both in your RAG…
The hardest retrieval problem in Graph RAG is not missing data. It is the query that arrives with an...
