RAG Vector Retrieval Workbench

D3 & Cosine Sim Engine
Benchmark Scenarios:
Retrieval Strategy
Dense Vector
Sparse BM25
RRF Hybrid
Top-K Bounds 3
Similarity Threshold 0.72
Chunk Size (chars) 60
Chunk Overlap 15
Search Query
2D Embedding Vector Space (Cosine Radii) D3 Projection Live
Query Vector
Retrieved Chunk
Unfiltered Chunk
Similarity Radius
Document Chunk Corpus
Grounding Inspector & Faithfulness Telemetry
Total Chunks 5
Retrieved 3
Mean Similarity 0.842
Faithfulness 94.5%
Synthesized Response vs. Context Grounding 2 Spans Verified
Retrieved Context Window
Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.