Grep or Embeddings? Agentic Search Over Company Documents — George He, LlamaIndex

תקציר מקורי באנגליתGrep works great on a codebase. Try it on millions of PDFs, slides and scanned schematics. George He, head of engineering at LlamaIndex, explains when agents need pre-indexed search and when plain file traversal is enough. Claude Code skipped embeddings because code is small, text-based, local and full of breadcrumbs. Company data is huge, messy, multimodal and permissioned. He shows how to give an agent a harness of tools so it can choose: hybrid retrieval as a compass, then listing, metadata filtering, grep and reading to dig in, plus parsed text and page screenshots for complex documents. He covers parsing with LlamaParse and the open-source LiteParse, permissions and freshness in production, and the ingest-parse-index pipeline. Then he demos an agent building a cash-flow table from 135