Bright Data — RAG / Search-Engine Pipeline

Use Discover as the retrieval layer for an LLM app or a custom search engine. Discover already returns intent-ranked, relevance-scored results with parsed page content, so it does the "search + fetch + clean" stage of RAG for you. This is a code/architecture skill built on the discover-api skill — read that for API mechanics (trigger/poll, modes, params, limits).

Pick the right neighbor: a written brief → live-research; markdown of specific URLs you already have → scrape; structured platform records → data-feeds.

Two architectures — choose first

Does the corpus change every query, or is it a stable knowledge base?