thu-vu92/graphRAG ? reverse-engineered prompt

Build me a notebook based project that searches the web for AI copyright and governance content, pulls in article text and YouTube transcripts, then turns that into a knowledge graph I can explore.

I want one notebook for collecting the data and another for building the graph and answering questions from it. The graph should extract entities and relationships from the scraped content, group related topics into communities, and create a clean interactive HTML view where I can search, filter by entity type, click nodes to see connections, and hover for details.

Please make it work end to end with a simple setup using API keys from a .env file, and include the exported CSV, graph JSON, and the final HTML visualization. If you need to check current library docs online while wiring it up, go for it.