LexTransition-AI

Live Demo: https://kvbgkvw4mehwhhdjt7crrg.streamlit.app/

LexTransition AI is an open-source, offline-first legal assistant. It helps users navigate the transition from old Indian laws (IPC/CrPC/IEA) to the new BNS/BNSS/BSA frameworks. Using local Machine Learning and OCR, it analyzes legal documents and maps law sections with 100% grounded accuracy.

---

⚖️ LexTransition AI: Law Mapper & Document Analyzer

LexTransition AI is an open-source, offline-first legal assistant. It helps users navigate the transition from old Indian laws (IPC/CrPC/IEA) to the new BNS/BNSS/BSA frameworks. Using local Machine Learning and OCR, it analyzes legal documents and maps law sections with 100% grounded accuracy.

---

🚀 Key Modules

🚀 Core Features

• 🔄 Intelligent Law Mapper: Maps old IPC sections to new BNS equivalents. Uses an LLM to highlight specific changes in wording, penalties, and scope.
• 🖼️ Multimodal OCR Analysis: Upload photos of legal notices or FIRs. The system extracts text using local OCR and generates actionable summaries.
• 📚 Grounded Fact-Checking (RAG): Ask legal questions and get answers backed by official citations. The AI identifies the exact Section and Page from uploaded Law PDFs to prevent hallucinations.
• 🎙️ Environment-Aware Voice Agent: Features high-fidelity offline TTS (Piper) with an automatic, lightweight cloud fallback (gTTS) to ensure seamless audio playback on headless platforms like Streamlit Cloud.

---

🛠️ Offline Tech Stack (No-API Approach)

To ensure privacy and offline accessibility, this project can be configured to run without external APIs:

• Frontend: Streamlit
• Backend: Python, LangChain/LlamaIndex.
• Local LLM Engine: Ollama (Llama 3 / Mistral)
• Voice / TTS: Piper TTS (ONNX models)
• OCR Engine: EasyOCR / PyTesseract
• Vector Database (RAG): FAISS + Sentence-Transformers

---

📂 Project Structure

LexTransition-AI/
├── .github/
│   └── workflows/
│       └── lextransition-ci.yml  # GitHub Actions CI/CD pipeline
├── engine/
│   ├── comparator.py             # AI logic for comparing IPC & BNS texts
│   ├── llm.py                    # Fallback logic and LLM summarization
│   ├── mapping_logic.py          # Core IPC to BNS transition logic
│   ├── ocr_processor.py          # Local OCR extraction and processing
│   ├── rag_engine.py             # Local Vector Search logic (FAISS)
│   └── db.py                     # Database connection and queries
├── utils/
│   └── timeout_handler.py        # Resiliency and API timeout handlers
├── tests/
│   └── test_embeddings.py        # Pytest suite for automated testing
├── scripts/
│   └── ocr_benchmark.py          # OCR character error rate testing
├── models/
│   └── tts/                      # Local storage for Piper ONNX voice models
├── law_pdfs/                     # Upload directory for Grounded Fact-Checking
├── app.py                        # Main Streamlit UI application
├── Dockerfile                    # Production container configuration
├── requirements.txt              # Python dependencies & OS-specific markers
├── setup_agent.py                # Manual setup script for downloading TTS binaries
└── README.md                     # Master project documentation

---

⚙️ Installation & Local Setup

Option A: Using Docker (Recommended)

The easiest way to run LexTransition-AI is with Docker. This handles all dependencies (including Tesseract OCR and system libraries) automatically.

1. Clone the repository:

   git clone [https://github.com/[username]/LexTransition-AI.git](https://github.com/[username]/LexTransition-AI.git)
   cd LexTransition-AI

2. Build the Docker Image in terminal

   docker build -t lextransition .

3. Run the Application

   docker run -p 8501:8501 -e LTA_OLLAMA_URL="[http://host.docker.internal:11434](http://host.docker.internal:11434)" lextransition-ai

4. Open the App

   http://localhost:8501

---

Option B: Manual Local Setup (Windows/Linux/Mac)

If you prefer to run the app directly in your local Python environment:

1. Install Dependencies (requires Python 3.10)

   python -m venv venv
   source venv/bin/activate  # On Windows use: venv\Scripts\activate
   pip install -r requirements.txt

2. Download Voice Agent Models

   python setup_agent.py

3. Start the Local LLM

   ollama serve
   ollama pull llama3

4. Launch the App

   export LTA_OLLAMA_URL="http://localhost:11434"  # On Windows use: set LTA_OLLAMA_URL=http://localhost:11434
   streamlit run app.py

---

🟢 Current Implementation Status & Architecture

All core modules, offline LLM integrations, and containerization features are fully implemented and production-ready.

   =========================================================================
                  🚀 LEXTRANSITION-AI: SYSTEM ARCHITECTURE
   =========================================================================

                  [ 🖥️ Streamlit Frontend (app.py) ]
                                 |
         -------------------------------------------------
         |                       |                       |
   [ 🔄 IPC → BNS Mapper ]  [ 🖼️ Document OCR ]   [ 📚 Fact-Checker (RAG) ]
         |                       |                       |
   (SQLite Mapping DB)    (EasyOCR / PyTesseract)  (FAISS + sentence-transformers)
         |                       |                       |
         -------------------------------------------------
                                 |
                                 v
                  [ 🧠 Local LLM Engine (Ollama) ]
            (Semantic Analysis, Action Items, Summarization)
                                 |
                                 v
                  [ 🎙️ Offline Voice Agent (Piper TTS) ]
            (High-fidelity vocal dictation of AI outputs)

   =========================================================================
                     ⚙️ INFRASTRUCTURE GUARANTEES
   =========================================================================
   ✔️ 100% Offline Capable (No external API keys required)
   ✔️ Dockerized Deployment (Verified networking & TTS dependencies)
   ✔️ CI/CD Pipeline Active (GitHub Actions + Pytest)

---

💾 Data Persistence & Testing

1. Local Data Storage (Privacy-First) To maintain our strict offline-first architecture, no user data or legal documents ever leave your machine:

• Relational Data: Mappings and system configurations are persisted securely using a local SQLite database (replacing the legacy mapping_db.json).
• Vector Store: Uploaded law PDFs for Grounded Fact-Checking are processed and stored locally in a FAISS vector index (./vector_store).

2. Automated Testing & CI/CD LexTransition-AI maintains high reliability through local testing and GitHub Actions.

   Local Unit Tests To run the test suite locally, ensure your virtual environment is active (Python 3.10) and execute:

   pip install -r requirements.txt
   pytest -q

---

Continuous Integration (GitHub Actions)

Every Pull Request automatically triggers our .github/workflows/lextransition-ci.yml pipeline.

---

OCR Benchmark Harness

To evaluate the local OCR engine's Character Error Rate (CER) and Keyword Recall against custom scanned datasets

python scripts/ocr_benchmark.py --dataset data/ocr_dataset.csv --report ocr_report.md

---

⚙️ Advanced Configuration (Environment Variables)

LexTransition-AI is designed to be plug-and-play, but power users can customize the engine behavior using environment variables. If you are using Docker, these are passed via the -e flag.

Variable	Default	Description
LTA_OLLAMA_URL	http://localhost:11434	The endpoint for the local LLM. When running in Docker, use http://host.docker.internal:11434 to route traffic to your host machine.
LTA_OLLAMA_MODEL	llama3	Specifies which local model to use for analysis and summarization.
LTA_USE_EMBEDDINGS	1	Toggles the FAISS/Sentence-Transformer RAG engine. Set to 0 to fallback to legacy keyword search.

---

🗺️ Project Roadmap & Future Scope

All foundational features (Local LLM, OCR, Vector DB, and CI/CD) are fully operational. The next phase of development focuses on expanding accessibility and enterprise utility:

• Speech-to-Text (STT) Integration: Implement local Whisper models to allow users to verbally query the Fact-Checker without typing.
• Multilingual Support (Indic Languages): Translate BNS mappings and OCR summaries into Hindi, Bengali, and other regional languages for broader accessibility.
• Precedent & Case Law Expansion: Expand the RAG Vector Database beyond standard Bare Acts to include landmark judicial precedents.
• Automated Legal Briefs: Add a reporting engine to export OCR analysis and IPC-to-BNS comparisons into cleanly formatted PDF/Docx files.

---

✨ Contributors

This project exists thanks to the amazing people who contribute their time, ideas, and improvements.

We truly appreciate every contribution 💙