| title | DataSentry v2 |
|---|---|
| emoji | 🔍 |
| colorFrom | blue |
| colorTo | green |
| sdk | gradio |
| sdk_version | 4.7.1 |
| app_file | app.py |
| pinned | false |
| python_version | 3.11 |
Hybrid PII/PHI detection engine with 4-layer architecture and full audit trail.
🔗 https://sriDevelops-datasentry.hf.space
- Layer 1: Regex patterns (SSN, email, credit card, ICD codes)
- Layer 2: spaCy NER (person names, locations, facilities)
- Layer 3: Claude LLM arbitration (ambiguous entities below 0.75 confidence)
- Layer 4: SQLite audit trail (every decision logged)
Python 3.11 · spaCy 3.7.4 · Claude API · FastAPI · Gradio
docs/DECISIONS.md— Architecture decisions with rationaledocs/METRICS.md— Measured recall, FP rate, latency, test coveragedocs/LIMITATIONS.md— What DataSentry does not do, and whydocs/COVERAGE.md— What flags standalone vs with context
- GitHub: https://github.com/sri-25/datasentry
- Live Demo: https://sriDevelops-datasentry.hf.space
- Author: https://linkedin.com/in/srijan24