Hectate is a gender-gated, AI-enforced digital safe space for women — secured by a triple-layer biometric verification engine, powered by real-time sentiment moderation, and anchored to a verified legal empowerment hub. If you are not verified female, the door does not open.
[ LIVENESS CHECK ] → [ BIOMETRIC MATCH ] → [ UPLOAD AADHAAR ] → [ OCR EXTRACTION ] → [ ✓ ACCESS GRANTED ]
Online spaces for women are routinely weaponised against them. Hectate is built on one principle: the entry gate itself is the safety mechanism. Not community rules. Not moderation queues. The architecture.
- 🔐 No man can register — the AI won't let them.
- 🛡 No fake persona survives the biometric check.
- ⚖️ Every user knows their legal rights, on demand.
- 🤝 Every advocate on the platform is verified.
| Layer | Library | Role |
|---|---|---|
| 🧠 AI Identity Verification Python Backend |
pytesseract |
OCR engine — extracts Aadhaar numbers, names & gender from uploaded documents in milliseconds |
MediaPipe |
3D face mesh tracking for liveness detection — defeats static photos and deepfakes | |
DeepFace + RetinaFace |
Dual biometric match — RetinaFace handles low-light detection; ArcFace/VGG-Face confirms identity with academic-grade accuracy | |
OpenCV + Pillow |
Real-time frame manipulation, image processing & environmental lighting analysis | |
| 🛡 Security & Moderation | vaderSentiment |
NLP sentiment analysis — flags toxic and harassing content before it reaches the community feed |
PyMuPDF (fitz) |
High-performance rendering and data extraction from Aadhaar PDF uploads | |
| ✨ Frontend Experience React / Vite |
face-api.js |
Client-side real-time face detection — zero server latency, instant "AI is watching" feedback |
Spline (@splinetool) |
Interactive 3D landing portal — premium security aesthetic, not a simple sign-up form | |
GSAP (GreenSock) |
Cinematic dashboard entrance animations and high-performance scroll-driven transitions |
┌─────────────────────────────────────────────────────────────────┐
│ HECTATE ENTRY PROTOCOL 2.1 │
└─────────────────────────────────────────────────────────────────┘
STEP 1 ──▶ Liveness Detection (MediaPipe)
68-point 3D face landmark tracking via front camera
Blink, turn, nod prompts defeat static photos & deepfakes
Live session token generated on pass
STEP 2 ──▶ Biometric Match (RetinaFace + ArcFace)
Selfie compared against document photo
Cosine similarity threshold: ≥ 0.68
Low-light fallback: RetinaFace re-processes under enhanced exposure
STEP 3 ──▶ Document Upload User uploads Aadhaar card (image or PDF) PyMuPDF extracts raw content from PDF uploads
STEP 4 ──▶ OCR Extraction (pytesseract)
Reads: Full Name · Aadhaar Number · Gender Field
Gender field MUST read "Female" — any other value = REJECTED
RESULT ──▶ All 4 stages pass → JWT issued → Dashboard unlocked
Any stage fails → Session terminated · No data retained
A curated repository of Indian laws protecting women — built for readability, not legal jargon.
- In-App PDF Viewer — read official government documents without leaving the secure Hectate environment
- Raw Preview Snippets — quick-read summaries before diving into full legislation
- Laws Covered:
- Protection of Women from Domestic Violence Act, 2005
- Sexual Harassment of Women at Workplace (POSH) Act, 2013
- Indian Penal Code sections on violence & harassment
- Dowry Prohibition Act, Maternity Benefit Act, and more
| Feature | Detail |
|---|---|
| Scope | Female advocates across all Indian states |
| Verification | Manual + document-backed credential check |
| Contact | Encrypted in-app messaging only |
| Search | Filter by state, specialisation, language |
Posts pass through vaderSentiment before publication. Toxicity is intercepted at the architecture level — not flagged after the damage is done. Compound scores below threshold trigger automatic hold for review.
Python >= 3.9
Node.js >= 18
Tesseract OCR (system install)# 1. Clone the repository
git clone https://github.com/your-org/hectate.git
cd hectate
# 2. Backend setup
cd backend
pip install -r requirements.txt
cp .env.example .env # Fill in your secrets
# 3. Frontend setup
cd ../frontend
npm install
# 4. Launch (two terminals)
# Terminal A:
cd backend && python app.py
# Terminal B:
cd frontend && npm run devHECTATE_SECRET_KEY=your_jwt_secret_here
OCR_ENGINE=tesseract
FACE_MODEL=ArcFace
LIVENESS_THRESHOLD=0.75
BIOMETRIC_SIMILARITY_THRESHOLD=0.68
SENTIMENT_TOXICITY_THRESHOLD=-0.5 ┌──────────────────────┐
│ React / Vite UI │
│ Spline · GSAP · PWA │
└──────────┬───────────┘
│ REST / WebSocket
┌──────────▼───────────┐
│ Python Flask API │
└──┬──────────────┬────┘
│ │
┌─────────────▼──┐ ┌──────▼──────────────┐
│ AI Auth Layer │ │ Moderation Layer │
│ │ │ │
│ pytesseract │ │ vaderSentiment │
│ MediaPipe │ │ Content Queue │
│ DeepFace │ │ Auto-flag + Hold │
│ RetinaFace │ └─────────────────────┘
│ OpenCV/Pillow │
└────────────────┘
| Principle | Implementation |
|---|---|
| Zero Trust Entry | Every session re-verified. No cookies bypass auth. |
| Data Minimisation | Biometric data processed in-memory, never persisted |
| Fail Closed | Any verification error = access denied, no fallback |
| Encrypted Transit | All API calls over HTTPS with JWT signing |
| AI Moderation First | Content never reaches feed without sentiment clearance |
"Safety is not a feature. It's the foundation."
