⚖️ LegalBro API (Backend)

LegalBro is an AI-powered legal assistant focused on Bangladeshi law and constitution.
It uses a Retrieval-Augmented Generation (RAG) system to provide accurate, context-aware answers from legal documents, simplified into plain English.

---

🚀 Features

• 🔍 AI-powered legal Q&A (RAG-based)
• 📚 Hybrid retrieval (Semantic + Keyword search)
• 🇧🇩 Focused on Bangladeshi law & constitution
• 🧠 Context-aware answers from legal PDFs
• ⚡ FastAPI backend with clean architecture
• 🔌 Pluggable LLM support (OpenAI / OpenRouter / DeepSeek)
• 📦 Preprocessed vector store (optimized for deployment)

---

🧠 How It Works

1. 📄 Legal PDFs are ingested and cleaned
2. ✂️ Text is chunked using structure-aware + contextual chunking
3. 🔗 Embeddings are generated using sentence-transformers
4. 🗂️ Stored in FAISS vector database
5. 🔎 Hybrid retrieval:
   • Semantic search (FAISS)
   • Keyword filtering (precision boost)
6. 🤖 LLM generates answers grounded in retrieved context

---

📁 Project Structure

---

app/ ├── rag/ │ ├── ingest/ # PDF loading, cleaning, chunking │ ├── retrieval/ # Hybrid retrieval logic │ ├── vectorstore/ # FAISS + metadata │ └── embedder.py ├── services/ # RAG + LLM orchestration ├── ai/ # LLM integration ├── api/ # API routes └── main.py # FastAPI entry point

data/ ├── processed/ # chunks.json └── vectorstore/ # index.faiss, metadata.pkl

---

⚙️ Setup

1. Clone repo

git clone https://github.com/fiction17/LegalBroBD.git

cd LegalBro

2. Create virtual environment

python -m venv .venv source .venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Environment variables

Create .env:

OPENAI_API_KEY=your_api_key

---

▶️ Run locally

uvicorn app.main:app --reload

Visit:

http://localhost:8000

---

🔗 API Endpoint

GET /ask

Example:

/ask?question=What is the basic structure of the Constitution of Bangladesh?

Response:

{
  "answer": "...",
  "sources": ["...", "..."]
}

🧠 RAG Pipeline

PDF → Clean → Chunk → Embed → FAISS
Query → Embed → Retrieve → LLM → Answer

🌍 Deployment

• Backend: Render
• Frontend: (Planned) Vercel

🛣️ Roadmap

• 🔥 Reranker (Cross-Encoder for better retrieval)
• 📄 Advanced legal chunking (section-aware)
• 🧾 Contract generation (e.g., partnership deed)
• 🧠 Multi-model support (DeepSeek, GPT, etc.)
• 🔐 Authentication & usage tracking
• ⚠️ Disclaimer

This project provides informational assistance only and does not constitute legal advice.

👨‍💻 Author

Built by Reyad