YIFU HE / INDEX00—4 / SHIPPED

SHIPPED / 2026 / Open-source Windows email quality tool

KOL DataCleaner

A local Windows tool that turns inconsistent creator exports into explainable outreach decisions — without sending a single email.

PythonPySide6pandasRule Engine
Data tunnel poster showing messy creator lists passing through validation, deduplication, scoring, and explainable routing.
ANNOTATION TRANSCRIPT
  • Keep every source column intact.
  • Extract and verify before scoring.
  • Explain every send, review, or skip route.
  • Catch agency context and shared contacts.
  • Local-first. It never sends email.
ROLE

Product definition, decision rules, data pipeline, desktop interface, testing, and Windows packaging.

STATUS

SHIPPED

OUTPUT

OPEN SOURCE / GITHUB

01 / WHY

DECIDE BEFORE YOU SEND.

PROBLEM

Creator exports arrive with inconsistent columns and several contacts mixed into free text. Valid-looking addresses can still be platform mailboxes, disposable accounts, shared agency contacts, or duplicates, so a send list assembled by eye is difficult to trust or audit.

SOLUTION

KOL Data Cleaner preserves every source column, extracts primary and secondary emails, then appends format, risk, optional MX, agency, duplicate, score, and routing fields to the same row. Each record ends in a clear send, review, or skip queue with a reason attached.

02 / SYSTEM

IMPORTVERIFYSCOREROUTE

  1. 01

    Merge one or more CSV and Excel exports, then confirm the automatically suggested field mapping.

  2. 02

    Extract every email candidate, normalize it, choose a primary address, and retain secondary contacts.

  3. 03

    Check format, risk type, optional MX, agency context, registry state, and duplicate relationships.

  4. 04

    Score each row, explain the decision, route it to send, review, or skip, and export one auditable file.

F—1

Batch import and merge for CSV, XLSX, XLSM, and XLS with source columns preserved.

F—2

Automatic field mapping with a manual confirmation layer for unfamiliar export schemas.

F—3

Primary and secondary email extraction from several mixed contact fields.

F—4

Format and risk validation with optional MX lookup, timeout handling, and a 30-day cache.

F—5

Agency detection, a local domain registry, and duplicate checks across email, creator, profile, and shared domains.

F—6

Explainable send, review, and skip routing with XLSX Summary and CSV export.

03 / EVIDENCE

BEFORE → BUILT → AFTER

BEFORE

Outreach readiness was judged by scanning inconsistent spreadsheets and reconstructing why an address looked safe or risky.

BUILT

A packaged Windows 0.1.0 desktop pipeline with local rules, optional cached MX checks, deterministic exports, and 28 passing automated tests.

AFTER

Every source row remains traceable while the appended score, reason, and final route make the next decision explicit. No real-list volume or outreach KPI is claimed.

KOL Data Cleaner fictional multi-file import and mapping view
Fictional CSV data demonstrates source-preserving import and mapping; no real creator records are shown.
KOL Data Cleaner fictional outreach decision results
Five fictional rows show send, review, duplicate, agency, and do-not-contact outcomes with reasons.
KOL Data Cleaner fictional agency registry and duplicate review
A fictional local registry records approved, contacted, pending-review, and do-not-contact agency domains.
LEARNING

A score alone is not enough. The useful output is a reversible decision: preserve the source, show the reason, and keep uncertain records out of the send queue until a person reviews them.

NEXT

Publish an explicit repository license, resolve the Fluent UI dependency before commercial distribution, version the rules and scoring model, and define a controlled export handoff into Pangdun Creator CRM.

SEE THE BUILD

[ GITHUB ↗ ]