This is managed RAG without the hassle of setting up a vector database. Upload PDFs, Word docs, Excel files, or 100+ other formats, and Google handles chunking, embeddings, and vector search automatically. The real win is document immutability forcing you into a delete-and-reupload pattern, which sounds annoying but prevents versioning chaos. Watch for the 3x storage multiplier (embeddings plus metadata) and only use Gemini 2.5 models. Citations come built in through grounding metadata. Best for knowledge bases where documents don't change hourly and you want predictable costs at $0.15 per million tokens indexed once. If you need real-time updates or custom embedding control, look at Cloudflare Vectorize instead.
npx skills add https://github.com/secondsky/claude-skills --skill google-gemini-file-search