LegalEase

LegalEase is a Streamlit-based AI legal assistant for Indian law. It helps users understand legal questions by retrieving relevant case-law context, mapping IPC sections to BNS sections, and surfacing relevant government schemes in a chat interfaces.

What it does

• Answers legal queries in English or Hinglish.
• Retrieves relevant text from case_data.csv using TF-IDF + cosine similarity.
• Detects IPC sections from the retrieved context and looks up the matching BNS section from mapping.csv.
• Suggests relevant government schemes when the user query matches legal aid, victim support, women safety, cybercrime, or other assistance topics.

Architecture

flowchart TD
    U[User] --> S[Streamlit UI: app.py]
    S --> B[backend.py]
    B --> R[TF-IDF Search over case_data.csv]
    R --> C[Relevant legal context]
    B --> M[IPC ↔ BNS mapping.csv]
    B --> G[Government schemes matcher]
    B --> L[Sarvam chat completion API]
    L --> B
    B --> S
    S --> U

Loading

Component flow

1. app.py renders the chat UI in Streamlit and sends user prompts to the backend.
2. backend.py builds the legal response.
3. case_data.csv is searched with TF-IDF and cosine similarity to find relevant legal text.
4. mapping.csv is used to translate IPC sections into BNS sections.
5. Sarvam API generates the final natural-language reply using the retrieved context.

How to run locally

1) Create and activate a virtual environment

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

On Windows:

.venv\Scripts\activate

2) Install dependencies

pip install -r requirements.txt

3) Make sure these files are present

• app.py
• backend.py
• schemes.json
• requirements.txt
• case_data.csv
• mapping.csv

4) Set up API Keys

Configure your Sarvam API keys using one of the following methods:

Method A: Streamlit Secrets (Recommended) Create a file at .streamlit/secrets.toml inside the project root directory and add your API keys:

SARVAM_API_KEYS = "key1,key2,key3"

Method B: Environment Variables Set the SARVAM_API_KEYS environment variable:

• Linux/macOS: export SARVAM_API_KEYS="key1,key2,key3"
• Windows (Cmd): set SARVAM_API_KEYS="key1,key2,key3"
• Windows (PowerShell): $env:SARVAM_API_KEYS="key1,key2,key3"

5) Run the app

streamlit run app.py

Demo steps

1. Open the app in your browser after running Streamlit.

2. Try one of the sample prompts on the home screen, such as:

   • What is the punishment for theft?
   • Domestic violence ke liye kya section hai?
   • Explain dowry laws in India
   • Mujhe arrest kiya bina warrant, kya karoon?

3. Watch the app retrieve legal context, map IPC to BNS, and generate a response.

4. Use the Dark Mode toggle in the sidebar to switch themes.

5. Click Start New Conversation to clear the chat and begin again.

Notes

• The app expects a working Sarvam API connection in backend.py.
• The legal responses are based on the uploaded dataset and mapping file, so the quality of case_data.csv and mapping.csv directly affects the output.
• This project is designed for educational and hackathon use.

Project structure

.
├── app.py
├── backend.py
├── schemes.json
├── requirements.txt
├── case_data.csv
├── mapping.csv
└── README.md

Security & Reliability Enhancements (June 2026)

The project has been refactored with the following improvements:

1. Security & API Keys: Hardcoded API keys in backend.py have been removed. Keys are now securely loaded dynamically from Streamlit Secrets (st.secrets) or standard environment variables (SARVAM_API_KEYS).
2. Correctness of IPC Section Extraction: The regex in backend.py was updated to require legal qualifiers (like "Section", "Sec", "IPC", "dhara") before extracting a section number, preventing false positives from matching arbitrary 3-digit numbers such as years (1860) or helpline numbers (1098).
3. Exact BNS Mapping Resolution: Refactored the mapping logic to search the structured dictionary metadata in the response column of mapping.csv rather than performing loose substring matching. This ensures queries for specific sections (like Section 37) don't return unrelated sections (like Section 376).
4. Government Schemes Database: Government schemes were extracted from the codebase into a clean, maintainable schemes.json configuration file.
5. Modern Theme Contrast: Polished Streamlit dark mode contrast dynamically, solving the problem of unreadable black text on light grey backgrounds for chat input fields in dark mode.