Privacy Shield

Detect, anonymize, and safely enhance text containing PII — locally, without sending sensitive data to anyone.

Privacy Shield is a single-container web app combining:

• FastAPI + Microsoft Presidio + spaCy for PII detection & anonymization
• A clean Vite + React + Tailwind SPA for three workflows:
  • Detect — find PII spans, types, and confidence scores
  • Anonymize — replace PII with realistic synthetic data
  • Enhance & reverse-anonymize — anonymize → call an LLM → restore original PII

Everything runs on your machine. The only outbound call is to your chosen LLM endpoint (and only if you use the Enhance tab).

Quick start (Docker)

docker run --rm -p 8000:8000 ghcr.io/senthilsweb/privacyshield:latest
# open http://localhost:8000

Or with compose (recommended for env management):

curl -L https://raw.githubusercontent.com/senthilsweb/privacyshield/main/docker-compose.yaml -o docker-compose.yaml
docker compose up -d

API

Method	Path	Purpose
GET	/health	Liveness probe ({"status":"ok"})
POST	/detect-pii-entities	Detect spans, types, scores
POST	/anonymize-with-fake-data	Replace PII with realistic fake values
POST	/anonymize-and-transform	Anonymize → LLM → reverse-anonymize
POST	/convert (opt-in)	Reserved for Markitdown; set ENABLE_CONVERT=true

Interactive Swagger UI is auto-published at /docs.

Configuration

Env var	Default	Notes
OPENAI_API_KEY	(unset)	Required only for the Enhance tab with OpenAI models
CORS_ORIGINS	*	Comma-separated origins
ENABLE_CONVERT	false	Mount the /convert placeholder route
LOG_LEVEL	INFO	DEBUG for verbose logs
PORT	8000	HTTP listen port

Architecture

flowchart LR
  user([Browser]) --> spa[Vite SPA<br/>3 tabs]
  spa -->|same-origin fetch| api[FastAPI]
  api --> presidio[Presidio + spaCy]
  api -. optional .-> llm[(LLM<br/>OpenAI / Ollama)]
  classDef ext fill:#fde68a,stroke:#a16207
  class llm ext

Loading

See docs/architecture.md for the full design and docs/about.md for the project background.

Development

Run locally without Docker

In dev the SPA and API run on two ports (Vite dev server can't serve a Python backend), but Vite reverse-proxies the API paths so the browser still sees a single origin. In the published Docker image, FastAPI serves both the built SPA and the API on one port (:8000).

# Backend — FastAPI on :8001
cd backend
python3.10 -m venv env && source env/bin/activate
pip install -r requirements.txt && pip install "uvicorn[standard]"
python -m spacy download en_core_web_lg
uvicorn api:app --reload --port 8001

# Frontend — Vite on :3000, proxies /detect-pii-entities, /anonymize-*,
# /entities, /prompts, /convert, /health → http://127.0.0.1:8001
cd ../frontend
npm install
npm run dev

Open http://localhost:3000. Hit the API directly at http://localhost:8001/docs for Swagger.

License

MIT © Senthilnathan Karuppaiah