ADVY is a Local-only data-privacy management system designed specifically to detect and protect personal data (PII) on the user's own machine. It puts security first with a self-contained design that never sends data outside unless you explicitly tell it to. A key strength is full Thai-language support through a 4-layer Multi-layer Detection engine that blends rule-based logic with AI for accurate, traceable results (Explainable Risk). ADVY's architecture spans risk scanning, PDPA compliance assessment, and policy enforcement that creates copies to avoid destroying the original files. It also includes an AI Gateway that screens data before it reaches large language models, ensuring confidential data never leaks in the age of working alongside AI.
Thailand's Personal Data Protection Act imposes civil, criminal, and administrative penalties for data that leaks or is retained without a lawful basis.
Payroll spreadsheets, scanned ID cards, and contracts often sit in shared folders that anyone can open, with no clear owner accountable for them.
Copying customer data into an AI tool without removing personal information first can push that data outside the organization in an instant.
ADVY stands for ARAK Data Discovery a tool for individuals and small teams who need to know where personal data lives, how risky it is, and how to send data to AI safely without relying on the cloud, standing up a server, or needing an IT team.
Walks every file in a folder to find personal data using Microsoft Presidio and PyThaiNLP covering Thai national IDs, bank account numbers, credit cards, names, emails, and credentials.
Every file receives a 0โ100 risk score with an explainable breakdown data sensitivity, access permissions, file age, and storage location.
Classifies and labels files automatically with a confidence level from PUBLIC to RESTRICTED, plus PDPA-PERSONAL, PCI-DATA, HEALTH-DATA, and CREDENTIAL.
Maps data to the 10 PDPA data categories and onward to 9 standards Cybersecurity Act, ISO 27001/27701, NIST CSF, PCI DSS, and SOC 2.
Supports quarantine, secure-copy, watermarking, and masking/redaction written as copies under ~/.advy/enforce/, never modifying or deleting the originals.
Applies Privacy-Enhancing Technologies PETs (Anonymize/DP/Synthetic/Federated) before sending to Claude, with a policy gate that blocks credentials and requires anonymization for PDPA-covered data.
Every step runs on your own machine, with no cloud connection unless you choose to send data yourself.
Specify the path to scan, e.g. /mnt/c/HR, then start the scan.
Reads each file to find entities in both Thai and English, and validates national ID checksums.
Combines risk factors into a 0โ100 score and a critical / high / medium / low rating.
Auto-labels files and maps them to PDPA data categories and the relevant standards.
Quarantine, redact, or send safely through the AI Gateway.
~/.advy/state.db, which persists across restarts.
The core ideas, the workflow, and the technology behind ADVY summed up in a few clear visuals.
The four core principles at the foundation of ADVY that ensure sensitive data never leaks through the tool itself.
A closer look at each step and the technology behind it. Every stage runs locally on your own machine.
/mnt/c/HR
ADVY automatically links the data found in each file to the relevant standards and laws, keeping a mapping record and audit log for verification.
Runs entirely on-device, with no cloud connectors and no telemetry so sensitive data never leaks through the tool itself.
Validates Thai national IDs with mod-11, detects Thai names via NER and bank account numbers, and understands Thai keywords for health, HR, and finance.
No server, no central database, and no IT team required just pip install and you're ready, even in regex-only mode.
Maps to 9 standards and laws, exports CSV/JSON, and keeps an enforce log evidence to support a PDPA audit.
A policy gate keeps sensitive data from leaking into prompts, so you get the benefit of AI without risking a PDPA breach.
Every action creates a new copy and never edits or deletes the source so you can experiment without harming your data.
Actual ADVY screens โ from the overview dashboard and data discovery to the catalog and compliance assessment.
Install the base build to run anywhere (regex-only), or add extras to enable full Thai and English NER.
C: drive lives at /mnt/c scan it directly, e.g. advy scan /mnt/c/Users/You/Documents