Visualizers
Learn by moving things
Interactive explainers for the ideas behind retrieval, embeddings, and agents. A formula tells you what is true; dragging a vector shows you why. Each one also appears inside the article it belongs to, and runs entirely in your browser, so drag, poke, and break it.
HyDE: search from a made-up answer
From: A Field Guide to RAG →A question sits far from the passages that answer it on the embedding map. HyDE drafts a rough made-up answer, which lands beside the real answer passages because it is shaped like one, and searches from there. The guess is discarded; the real passages it found are read.
The question is worded nothing like the passages that answer it, so it sits far away on the map and pulls back weak matches.
Contextual compression
From: A Field Guide to RAG →A retrieved chunk is on-topic but mostly padding for any one question. Contextual compression keeps only the sentences that actually answer it and drops the rest, so the prompt fills with signal instead of filler.
The retrieved chunk is on-topic but mostly padding for this exact question. Sending all of it wastes prompt budget.
Keeping the catalog in step with the shelf
From: RAGOps: RAG Is a Distributed System, Not a Demo →A source page is edited. Instead of rebuilding the whole index, a file watcher fires and only that one page is re-chunked and re-embedded (8 chunks, not 50,000), so the index catches up to the corpus one document at a time.
The index is built once. The policy page and its 8 chunks in the index agree: both v1, fresh.