Built for the Indian Servers AI Contest 2026
LexAI is a Retrieval-Augmented Generation (RAG) system that answers legal and tax queries with accurate, citation-grounded responses drawn from real legal corpora β no hallucinations, no guesswork.
- π Semantic Search using FAISS vector index + sentence-transformers (
all-MiniLM-L6-v2) - π€ AI Answers powered by Google Gemini API
- π Citation-aware responses β every answer references the exact legal section
- π Multi-corpus support β query across multiple legal documents simultaneously
- π Clean Web UI β Flask backend + Vanilla HTML/CSS/JS frontend
| Document | Jurisdiction |
|---|---|
| GST Act 2017 | India |
| Income Tax Act 1961 | India |
| IFRS 16 | International |
| UAE Corporate Tax Law 2023 | UAE |
| Layer | Technology |
|---|---|
| AI Model | Google Gemini API |
| Embeddings | sentence-transformers all-MiniLM-L6-v2 |
| Vector Store | FAISS |
| Backend | Flask (Python) |
| Frontend | HTML Β· CSS Β· Vanilla JavaScript |
User Query
β
βΌ
Sentence Embedding (all-MiniLM-L6-v2)
β
βΌ
FAISS Similarity Search β Top-K Legal Chunks
β
βΌ
Google Gemini API (RAG Prompt + Context)
β
βΌ
Citation-Aware Legal Answer
git clone https://github.com/Prabhas125/LexAI.git
cd LexAI
pip install -r requirements.txt
# Add your GEMINI_API_KEY in .env
python app.pyBuilt for the Indian Servers AI Contest 2026 β a competition focused on building AI-powered real-world applications.
MIT License