🎬 RAG Search Engine – Interactive Demo
Explore the progression from keyword → semantic → hybrid (RRF) → RAG on a 5 000-movie dataset.
Built from the open-source rag-search-engine CLI project.
Combines BM25 keyword and semantic embeddings with Reciprocal Rank Fusion (RRF).
1 15
Try these
Retrieve relevant movies with hybrid search, then generate a grounded answer using Gemini. Citations reference the retrieved documents.
1 10
Try these
Upload a movie poster or still. CLIP embeds the image and finds movies whose title+description are closest in the joint embedding space.
1 10
Models: all-MiniLM-L6-v2 · clip-ViT-B-32 · gemini-2.5-flash-lite
Note: First request after cold start downloads models and builds embeddings (can take 1–2 min).
Source: github.com/Utkarsh736/rag-search-engine