Web pages as ready-to-embed chunks for your vector store (template + how it works)

I built a small template that turns a web page into chunks you can embed directly, using only built-in nodes (HTTP Request + Split Out).

Why: fetching a page with HTTP Request and passing it on gives the model mostly navigation, scripts and cookie banners, and splitting by fixed length cuts sections in half. The template asks an API for the page split along its headings instead. Each chunk comes with a stable UUID (same page + position = same id, so a re-run overwrites instead of duplicating), the heading path, a token count and a content hash you can use to re-embed only what changed.

How it works:

Set node with the URL and max chunk size
HTTP Request to /v2/scrape with “formats”: [{“type”: “chunks”, “maxTokens”: 500}], key in a Header Auth credential
Split Out on data.chunks → one item per chunk, ready for your embeddings + Qdrant/Pinecone/Supabase/pgvector nodes

Template: Split web pages into vector-ready chunks with Ragfeed | n8n workflow template

Disclosure: the API is Ragfeed, which I built and run (solo, Germany). It’s Firecrawl-compatible, so the official Firecrawl node works too if you set its Base URL to https://api.ragfeed.de/v2. Free tier is 1,000 pages a month. Happy to hear how you handle updates of pages that change, that’s the part I’m least sure about for typical n8n setups.