Getting started with Smart Dedupe

The first useful result is not a cleaner vault. It is one candidate pair you understand well enough to keep, consolidate, archive, or ignore deliberately.

Before you start

Confirm that:

• Smart Dedupe Pro is installed and active
• the note you want to review is open
• Smart Environment has prepared blocks when you want semantic matches
• the first note contains material you can evaluate safely

Start with a normal working note, not customer data, credentials, legal records, or another high-risk source.

1. Open the Duplicate Detector

Open the Command Palette. Run this command:

Smart Dedupe Pro: Open Duplicate Detector

The persistent Duplicate Detector keeps scope, controls, progress, and candidates in one view.

Choose Current note for the first pass. Use Full vault only after the controls and review process feel predictable.

2. Set a conservative first boundary

For the first current-note scan:

1. Keep exact matches visible.
2. Set Max results to a number you can review in one session.
3. Use enough Minimum length to avoid headings and boilerplate.
4. Exclude frontmatter when repeated YAML is not useful.
5. Start around 0.90 for a strict review. Lower toward 0.85 when you intentionally want paraphrases.

The current controls may open with a different saved value. Review the visible setting instead of assuming it is a universal default.

See Configure the scan for every control.

3. Run the note scan

Choose Run note scan. Let the Detector settle.

A short scan can complete before the running state is easy to read. A longer scan shows progress. It also shows a Cancel action. If you cancel, wait for the status to settle. Treat any remaining candidates as a partial set.

4. Review one candidate pair

For one promising result, check:

1. exact or semantic match type
2. similarity value
3. Source and Duplicate note identities
4. headings or line ranges
5. surrounding text in both notes

Choose Open when the excerpt does not explain why each passage exists. Use Copy only when you deliberately need the text for comparison. Copy does not merge anything.

Source and Duplicate name the two sides of the interface. Neither label identifies the canonical note.

5. Make one human decision

Choose the outcome that matches the notes' roles:

• Keep both when the passages serve different audiences, projects, or stages.
• Clarify the distinction when both notes are useful but too easy to confuse.
• Consolidate manually when one source should absorb the useful material.
• Archive one source when it is genuinely redundant but you want a reversible cleanup path.
• Ignore the candidate when the overlap is harmless.

Smart Dedupe does not perform this decision for you.

You know it worked when

• the Detector returned at least one understandable candidate, or a clear zero-result state
• you inspected both sources rather than relying on the score
• you made one explicit keep, clarify, consolidate, archive, or ignore decision
• any note change was deliberate and reversible

If the first scan is weak

Continue from here

• Understand every Dedupe control
• Review candidates and result limits
• Use Connections for related, non-duplicate material
• Rebuild reviewed context after cleanup