Law Exam Batch Processor

Enterprise‑grade, self‑hosted system for generating strict, exam‑oriented Indian law answers, performing domain‑prioritized fact‑checking, and exporting verified PDFs.

1. Overview

Law Exam Batch Processor is a production‑ready Flask application designed for Indian law students, researchers, and legal professionals who require:

Strict, marks‑oriented exam answers
Minimal but authoritative case‑law usage
Statute‑first accuracy with verification
Batch processing of questions
Single consolidated, printable PDF output

The system combines:

Local LLM inference (Ollama)
Self‑hosted metasearch (SearXNG)
Deterministic prompt control
Pandoc + wkhtmltopdf publishing

2. Architecture

Backend

Flask – REST API and task orchestration
Threaded workers – Non‑blocking batch execution
Ollama – Local LLM inference (Gemma 3 4B)
SearXNG – Fact‑checking via Indian legal domains
Pandoc + wkhtmltopdf – Markdown → PDF pipeline

Frontend

Pure HTML + CSS + JavaScript (no framework)
Dark, distraction‑free exam interface
Live progress polling
Incremental answer rendering
One‑click final PDF export

3. Key Features

Strict exam‑safe answer format
One‑case‑law rule enforced
Statute‑priority verification
Domain‑weighted search ranking
Per‑question Markdown archival
Final combined verified PDF
Self‑hosted, offline‑friendly
No third‑party APIs or data leakage

4. Answer Policy (Exam Mode)

Each answer strictly follows:

Meaning / Direct Answer
Statutory Provision
Essential Points (brief explanation only)
Case Law (ONE most relevant case)
Conclusion
Confidence Level

Explicitly excluded:

Academic discussion
Multiple case laws
Illustrations
Comparative commentary

5. Fact‑Checking Strategy

Queries routed through self‑hosted SearXNG
Domain‑priority scoring
Preferred sources:
- indiankanoon.org
- sci.gov.in
- supremecourtofindia.nic.in
- highcourts.gov.in
- gov.in
LLM instructed to:
- Correct errors only
- Preserve structure
- Avoid expansion

6. Directory Structure

project-root/
│
├── app.py              # Flask backend
├── templates/
│   └── index.html      # Frontend UI
├── md/                 # Per‑question markdown
├── output/
│   ├── *_final.md
│   └── *_final.pdf
└── README.md

7. Requirements

System

Linux (recommended)
Python 3.10+
8 GB RAM minimum (16 GB recommended)

Dependencies

Flask
requests
ollama (local runtime)
pandoc
wkhtmltopdf
SearXNG (self‑hosted)

8. Installation

1. Clone Repository

git clone <your-repo-url>
cd law-exam-batch-processor

2. Python Environment

python3 -m venv venv
source venv/bin/activate
pip install flask requests ollama

3. Install System Tools

sudo apt install pandoc wkhtmltopdf

4. Setup Ollama

ollama pull gemma3:4b

5. Setup SearXNG

Deploy SearXNG (Docker recommended)
Ensure /search?format=json is accessible
Update SEARXNG_BASE_URL in app.py

9. Running the Application

python app.py

Access UI at:

http://localhost:5000

10. API Endpoints

Endpoint	Method	Purpose
/exam	POST	Submit batch questions
/progress/	GET	Poll progress
/download/pdf/	GET	Download final PDF

11. Security & Privacy

Fully self‑hosted
No cloud inference
No third‑party data sharing
Local network deployment possible
Suitable for confidential exam preparation

12. Intended Use

Law examinations
Judicial service preparation
University assessments
Legal revision notes
Offline legal research

Not intended for:

Casual Q&A
Chat‑style responses
Opinion‑based analysis

13. License

This project is released under a permissive open‑source license.

You are free to:

Use
Modify
Self‑host
Deploy commercially

Attribution is appreciated.

14. Credits & Acknowledgements

This project builds upon and credits the following open‑source software and platforms:

Python – Core language runtime
Flask – Web framework
Ollama – Local LLM serving
Gemma Models – Google DeepMind
SearXNG – Privacy‑respecting metasearch engine
Pandoc – Universal document converter
wkhtmltopdf – HTML/Markdown to PDF rendering
Indian Kanoon – Public legal information (referenced, not scraped)

All trademarks and copyrights belong to their respective owners.

15. Disclaimer

This software is provided for educational and research purposes.

While best efforts are made to ensure legal accuracy, users must independently verify answers before reliance in professional or judicial contexts.

16. Maintainer

Maintained by an independent legal‑tech developer.

Contributions, audits, and improvements are welcome via pull requests.

Code Link

View Original Repository on GitHub ↗

← Back to Projects