Hey n8n Community,
Document classification in n8n is one of those things that looks complicated until you realize how little setup it actually needs. With the easybits Extractor it’s a 2-node workflow and a single field — and if you want to extract other data from the same document in the same pass, you just add more fields. I recorded a short walkthrough of the full setup.
Video: Document Classification in n8n – The Easy Way (Full Setup Guide)
The whole thing is two nodes: a form trigger to accept a file upload, and the easybits Extractor node with a single document_class field. The classification prompt lives in that field’s description – it tells the model which categories to choose from and to return null if nothing fits. That’s it. No separate classifier node, no chain of prompts, no HTTP request node.
What’s in the video:
- Setting up an easybits pipeline from scratch with a single classification field
- How to adapt the classification prompt to your own document types
- Installing the community node in n8n
- Wiring it up to a form trigger and running two test documents through it
Setup recap
- Cloud users:
easybits Extractoris available out of the box, search for it in the node panel - Self-hosted: Settings → Community Nodes → install
@easybits/n8n-nodes-extractor
Free tier is 50 requests/month, enough to test this end-to-end.
Want the production-ready version?
The video keeps things minimal on purpose – two nodes, one field, just to show the core pattern. If you want the version I actually run, it adds a second field for confidence_score and an IF node that routes empty or low-confidence results to Slack for manual review. Workflow JSON, both prompts, and the setup guide all sit in one GitHub folder:
Anyone else doing classification this way, or are you running it through a separate classifier node? Curious whether this pattern has made it further than I think.
Best,
Felix
1 Like