Home▸AI & Machine Learning▸RAG at Scale: Vector Search & Hybrid Retrieval Simulator

🔍 RAG at Scale: Vector Search & Hybrid Retrieval Simulator

Index thousands of chunk embeddings in a live 3D vector space, tune chunk size and top-K, toggle hybrid BM25+vector reranking, and watch latency and recall trade off in real time.

AI & Machine Learning3DModerate60 FPS

Details

CategoryAI & Machine Learning

RenderingThree.js · WebGL

DifficultyModerate

Target FPS60

ControlsChunk-size buttons (Small/Medium/Large) change how tightly each topic cluster is packed; Indexed-chunks slider (200-3000) scales the vector index and drives search latency; Top-K slider (1-24) sets how many nearest neighbors are retrieved and connected to the query; Hybrid BM25+vector toggle blends a simulated keyword score into the ranking; New Query button samples a fresh query vector and re-runs retrieval instantly.

Three.jsRAGvector searchembeddingsretrievalInstancedMesh

Related simulations

Agent Memory: Retrieval & ConsolidationWatch episodic memories decay, get pulled by hybrid retrieval, consolidate into semantic…

Conversational AI: Chatbot Dialog Manager & RAG SimulatorInteractive 3D simulation of a chatbot's architecture: a dialog manager holding the…

Long-Context Retrieval SimulatorWatch chunked documents live in embedding space, queries pull the top-K nearest chunks by… ⊞ All 880 in AI & Machine Learning →

⚙ Under the hood

Index thousands of chunk embeddings in a live 3D vector space, tune chunk size and top-K, toggle hybrid BM25+vector reranking, and watch latency and recall trade off in real time.

Three.jsRAGvector searchembeddingsretrievalInstancedMesh

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install