Turn any input (code, docs, papers, images, videos) into a navigable knowledge graph, then answer questions against it.
