Legal Sahayak: RAG‑Based Explainable Legal Copilot for India

Note

This project was developed as part of EPICS (Engineering Projects in Community Service), an initiative to apply engineering skills for social good and community empowerment.

❗ Problem Statement

Every day, millions of common people across India face situations requiring timely legal advice—ranging from landlord disputes and consumer rights to family law and labor issues. However:

1. Immediate Help Unavailable

Legal aid clinics and lawyers are often inaccessible due to cost, geography, or language barriers. In rural Madhya Pradesh, the lawyer-to-population ratio is approximately 1:10,000 compared to the national average of 1:1,200, leaving vast populations without legal representation.

2. Complexity of Legal Texts

National and state laws are voluminous, jargon‑laden, and constantly updated, making manual search slow and error‑prone. The Bharatiya Nyaya Sanhita (BNS) 2023 alone contains 358 sections, while consumer protection, labor, and family laws add thousands more provisions.

3. Contextual Relevance Missing

A generic search may surface outdated or non‑applicable statutes. What matters most are:

Local regulations (state-specific amendments)
Recent amendments (BNS 2023, BNSS 2023)
User‑specific factors (state, language, socio‑economic profile)

4. Lack of Explainability

Even when AI systems retrieve relevant sections, users can't trace "why" a particular clause applies or how it led to a recommended action plan. This black‑box approach erodes trust among vulnerable populations.

5. Digital Divide & Literacy Barriers

65% of Madhya Pradesh's population lives in rural areas
70.6% literacy rate (vs. 77.7% national average)
Limited smartphone penetration and poor internet connectivity in tribal districts like Alirajpur, Barwani, and Jhabua

🚀 Our Solution

Legal Sahayak is an AI-powered, explainable legal assistance platform that democratizes access to legal information through an integrated system combining intelligent search, personalized explanations, and a comprehensive digital legal library.

1. Intelligent Legal Search & Retrieval

FAISS Vector Store: Semantic search across 6+ domains of Indian law
Hybrid Retrieval: Combines dense embeddings (HuggingFace Transformers) with keyword matching
Contextual Chunking: Markdown-based hierarchical splitting preserves legal structure (Act → Chapter → Section)
Cross-references: Automatically links related laws and provisions
Metadata Preservation: Tracks source documents and specific legal sections

2. Demographic-Aware Personalization

Adapts explanations based on user profile:

Age & Education → Simplifies legal language for users with lower literacy
Location → Prioritizes MP-specific laws and local court procedures
Occupation → Provides job-relevant examples (e.g., labor laws for factory workers)
Gender → Ensures culturally sensitive responses for sensitive issues

3. Explainable AI Architecture

Every response includes structured information:

Plain-language Answer: Legal explanation in simple terms
Confidence Score: 0.9-1.0 (clearly defined), 0.6-0.8 (requires interpretation), <0.6 (recommends lawyer)
Legal References: Direct citations to applicable laws and sections
Action Plans: Step-by-step guidance with timelines, relevant authorities, required documents, and cost estimates

4. Bilingual Support (Hindi + English)

Full application interface in Hindi and English
Responsive language toggle for seamless switching
Culturally relevant explanations adapted to local context
Supports semi-literate users with simplified language options

5. Legal Library

Interactive, searchable database of Indian laws
Covers 6+ legal domains: Consumer Protection, Fundamental Rights, Marriage Laws (Hindu & Muslim), Labor Laws, Rental Laws
Real-time document browsing with hierarchical navigation
Quick reference guides for common legal scenarios

6. Privacy-First Design

No storage of personal legal queries
Demographic data used only for response generation
GDPR-compliant data handling

🔄 Query Processing Flow (Architecture)

sequenceDiagram
    participant U as User
    participant F as Frontend
    participant API as FastAPI Backend
    participant RAG as RAG Service
    participant FAISS as Vector Store
    participant LLM as Groq LLM

    U->>F: Enter legal question + demographics
    F->>API: POST /api/v1/qa/ask
    Note over F,API: {question, demographics}

    API->>RAG: Process QA Request

    RAG->>FAISS: Semantic Search (k=5)
    Note over RAG,FAISS: Convert query to embeddings
    FAISS-->>RAG: Top 5 relevant documents

    RAG->>RAG: Format context + demographics
    RAG->>LLM: Invoke with structured prompt
    Note over RAG,LLM: JSON Schema: QAResponseModel

    LLM-->>RAG: JSON Response
    Note over LLM,RAG: {answer, confidence,<br/>legal_references,<br/>action_plan}

    RAG->>RAG: Validate & Parse Response
    RAG-->>API: QAResponseModel
    API-->>F: HTTP 200 + JSON
    F-->>U: Display formatted answer
    Note over F,U: (1) Explanation<br/>(2) Confidence Score<br/>(3) Sources<br/>(4) Action Plan

🛠️ Technology Stack

Backend

Framework: FastAPI 0.115+ (async, high-performance)
LLM: Groq API (llama-3.3-70b-versatile or gpt-oss-120b)
Embeddings: HuggingFace sentence-transformers/all-MiniLM-L6-v2
Vector DB: FAISS (Facebook AI Similarity Search)
RAG Framework: LangChain 0.3+
Schema Validation: Pydantic v2

Frontend

Framework: Next.js 15 (App Router)
UI Library: shadcn/ui + Radix UI
Styling: Tailwind CSS
State Management: React Hooks
Markdown Rendering: react-markdown + remark-gfm

Data Pipeline

Document Format: Markdown (.md)
Text Splitting: LangChain MarkdownHeaderTextSplitter
Indexing: Python script (build_vectorstore.py)

📦 Installation & Setup

Prerequisites

Python: 3.11+
Node.js: 18+
pnpm: 8+ (or npm/yarn)
Git: For cloning the repository

1. Clone the Repository

git clone https://github.com/sujeetgund/legalsahayak.git
cd legalsahayak

2. Backend Setup

a. Create Virtual Environment

cd backend
python -m venv venv

# Windows
venv\Scripts\activate

# macOS/Linux
source venv/bin/activate

b. Install Dependencies

pip install -r requirements.txt

c. Configure Environment Variables

Create a .env file in the backend/ directory:

# API Keys
GROQ_API_KEY=your_groq_api_key_here
GROQ_MODEL_NAME=llama-3.3-70b-versatile

# Embedding Model
EMBEDDING_MODEL_NAME=sentence-transformers/all-MiniLM-L6-v2

# Security
BEARER_TOKEN=your_secure_bearer_token

# Paths
DATA_DIR=data
FAISS_INDEX_DIR=faiss_index

d. Build Vector Store (optional)

python build_vectorstore.py

This will:

Load documents from data/ directory
Split into hierarchical chunks
Generate embeddings
Create FAISS index in faiss_index/

e. Run Backend Server

uvicorn main:app --reload --host 0.0.0.0 --port 8000

Backend will be available at: http://localhost:8000

API docs: http://localhost:8000/docs

3. Frontend Setup

a. Install Dependencies

cd ../frontend
pnpm install

b. Configure Environment Variables

Create a .env.local file in the frontend/ directory:

BACKEND_SERVER_URL=http://localhost:8000/api/v1

c. Run Development Server

pnpm dev

Frontend will be available at: http://localhost:3000

4. Docker Deployment (Optional)

Full Stack with Docker Compose

# From root directory
docker-compose up --build

Services:

Frontend: http://localhost:3000
Backend: http://localhost:8000

🎯 Usage

Web Interface

Navigate to Chat Interface: http://localhost:3000
Fill User Profile (one-time):
- Age: 25-35
- Gender: Male/Female/Other
- Location: Bhopal, Madhya Pradesh
- Education: Graduate
- Occupation: Private Employee

Ask Your Question:

Example: "My landlord is asking me to vacate without notice. What are my rights?"

Review Response:
- Explanation: Plain-language answer
- Confidence Score: 85%
- Legal References: MP Rent Control Act, 1961 - Section 12
- Action Plan:
  1. Check your rent agreement
  2. Send legal notice with 1-month period
  3. File complaint if harassment continues

API Usage

Endpoint: `POST /api/v1/qa/ask`

Request:

{
  "question": "What is the minimum wage in Madhya Pradesh for unskilled workers?",
  "demographics": {
    "age": 30,
    "gender": "Male",
    "location": "Indore, Madhya Pradesh",
    "education_level": "High School",
    "job_title": "Factory Worker"
  }
}

Response:

{
  "answer": "As of 2024, the minimum wage for unskilled workers in Madhya Pradesh is ₹320 per day (₹9,600 per month for 30 days). This applies to industries covered under the Minimum Wages Act, 1948...",
  "confidence": 0.92,
  "legal_references": [
    "Minimum Wages Act, 1948 - Section 3",
    "MP Minimum Wages Notification 2024"
  ],
  "action_plan": [
    {
      "title": "Verify Your Wage Slip",
      "description": "Check if your employer is paying at least ₹320/day. If not, collect 3 months of wage slips as evidence."
    },
    {
      "title": "File Complaint with Labour Commissioner",
      "description": "Visit the nearest Labour Office (Office of the Commissioner, Labour Department, Indore). No fee required."
    },
    {
      "title": "Contact MP Labour Helpline",
      "description": "Call 155214 for immediate assistance and guidance on next steps."
    }
  ]
}

cURL Example

curl -X POST http://localhost:8000/api/v1/qa/ask \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer your_bearer_token" \
  -d '{
    "question": "How do I file a consumer complaint?",
    "demographics": {
      "age": 28,
      "gender": "Female",
      "location": "Bhopal, MP",
      "education_level": "Graduate",
      "job_title": "Teacher"
    }
  }'

📄 License

This project is licensed under the MIT License.

Name		Name	Last commit message	Last commit date
Latest commit History 14 Commits
backend		backend
frontend		frontend
README.md		README.md

Folders and files

Latest commit

History

Repository files navigation