Deduplicate and reconcile smart notes with Gemini, Supabase and TypeSafe Jev

Quick overview

This workflow receives notes via a webhook, splits them into individual points, and uses TypeSafe Jev to classify each point as new, duplicate, or an update to an existing note before storing the cleaned note ledger in n8n workflow static data and returning a JSON summary.

How it works

1. Receives a POST webhook request containing a note in JSON (for example in a note, text, or content field, with an optional title).
2. Splits the note into up to 20 cleaned, line-based points and loads the most recent saved notes from the workflow’s static data store.
3. Sends the incoming points and the saved note context to the TypeSafe Jev API to classify each point as new_information, duplicate, or an update_N targeting a specific saved note.
4. Applies Jev’s decisions by adding new points to the saved note ledger, appending updates to the matched existing notes, and skipping duplicates while collecting any ambiguous items for manual review.
5. Returns a JSON response with a summary of what was added, updated, skipped as duplicates, and flagged for review.

Setup

1. Add your TypeSafe Jev API key by replacing the Authorization: Bearer REPLACE_WITH_YOUR_TYPESAFE_JEV_API_KEY header value in the HTTP request.
2. Activate the workflow and copy the production webhook URL from the webhook trigger, then configure your notes source to send POST requests to it.
3. Send notes with one point per line (max 20 lines) in the JSON body using note (or text/content), and optionally include a title field.