v1.0.0

🚀 Unveil v1.0.0 — Initial Release

First public release of Unveil, an AI-powered fairness and explainability platform focused on uncovering hidden bias, improving transparency, and enabling responsible AI analysis. Built for the Google AI Solution Challenge using Gemini 2.5 Flash.

---

✨ Features

🧠 Fairness & Bias Analysis

• Bias detection workflows
• Slice-based fairness evaluation
• Approval gap analysis
• Disparate impact / four-fifths rule evaluation
• Group comparison metrics

🔍 Explainability

• SHAP-based feature explanations
• Model interpretability visualizations
• Transparent prediction analysis
• Black-box counterfactual probing

⚖️ Responsible AI Tooling

• Counterfactual analysis
• Fairness-aware reporting
• Explainability pipelines
• Proxy attribute detection
• Compliance-oriented AI summaries

⚙️ Platform & Infrastructure

• Modular backend architecture
• Interactive frontend dashboard
• Firebase integration
• Structured API pipeline

---

🛠 Tech Stack

Frontend

• React 19
• Vite
• TailwindCSS v4
• Framer Motion
• Recharts

Backend

• Python
• FastAPI
• Uvicorn
• Scikit-learn
• Pandas
• NumPy
• SciPy

AI & Explainability

• Gemini 2.5 Flash
• SHAP
• Fairness evaluation workflows

Infrastructure

• Firebase Auth
• Firestore
• Environment-based configuration

---

🌐 Live Demo

Frontend

https://unveil-201cc.web.app/

Backend API Docs

https://unveil-899475904423.europe-west1.run.app/docs

---

📦 Installation

1. Clone the Repository

git clone https://github.com/Jay-Jay-Tee/unveil.git
cd unveil

2. Environment Variables

cp .env.example .env

Configure at minimum:

• VITE_GEMINI_API_KEY — get one free at [aistudio.google.com](https://aistudio.google.com/app/apikey)
• GEMINI_API_KEY — same key, used by the backend
• Firebase vars (optional — app falls back to localStorage without them)

3. Start the App

The startup scripts install dependencies and launch both services together.

macOS / Linux:

chmod +x setup/start.sh
./setup/start.sh

Windows (CMD):

setup\start.bat

Windows (PowerShell):

.\setup\start.ps1

Or manually:

npm install
pip install -r docs/requirements.txt

# Terminal 1
npm run frontend

# Terminal 2
npm run backend

Services:

• Frontend → http://localhost:5173
• Backend → http://localhost:8001

---

🧪 Demo Workflow

1. Upload a dataset (CSV, XLSX, JSON, TSV, etc.)
2. Optionally upload a trained .pkl ML model
3. Unveil automatically:
   • identifies sensitive attributes
   • detects proxy variables
   • evaluates fairness metrics
   • generates SHAP explanations
   • performs counterfactual analysis
4. Receive:
   • fairness dashboards
   • explainability visualizations
   • AI-generated compliance narratives

---

📌 Highlights

• Detects hidden proxy bias in datasets
• Generates plain-English compliance reports
• Combines explainability + fairness workflows
• Supports black-box model auditing
• Designed for transparency-focused ML analysis
• Structured for extensibility and future research

---

⚠️ Note

This release is an initial research and hackathon version intended for experimentation, evaluation, and responsible AI exploration.