De-Scramble

De-Scramble is an open-source probabilistic identity-resolution CLI tool, Apache-2.0 licensed, stewarded by Deluge Limited.

De-Scramble resolves duplicate and fragmented person records into deduplicated golden records using probabilistic record linkage (Splink / Fellegi-Sunter) into an Apache Iceberg lakehouse.

View the repository

Verified performance

On the demo dataset: 3,149 raw records → 2,009 golden records; 574 duplicate clusters resolved; 99.91% of candidate pairs eliminated by blocking.

Against known ground truth: precision 0.989, recall 0.984, F1 0.987.

These metrics reproduce identically from a clean clone of the public repo.

Engineering integrity

De-Scramble was built from scratch as an information island — no private source or history in the public repo; commit history masked; trace-scan clean across all blobs.
CI is green across a 5 OS/Python matrix, and a documented igraph-free (GPL-free) install path is CI-enforced.
License compliance is exact: De-Scramble is an Apache-2.0 project, and the one GPL transitive dependency (igraph via Splink) is kept off the documented install path, CI-verified, and precisely disclosed in NOTICE.
Cross-platform setup is documented (Windows PowerShell/cmd/Git Bash, macOS, Linux), with a visible report→fix→close maintenance cycle on the public repo.

The honest positioning

De-Scramble is genuine open source built to institutional standards: verification-first, compliance-by-design, reproducible.

The public repo is https://github.com/De-Scramble/descramble.