🧠Legal-Advisor-using-gpt-neo-1.3B

An intelligent legal assistant that leverages Natural Language Processing (NLP), vector search, and open-source language models to provide legal guidance based on authoritative legal texts.

---

🔍 How It Works

1. 💬 User Input: The user submits a legal question or prompt through a simple interface.
2. 📚 Semantic Search: The prompt is embedded and used to query a vector database containing semantic representations (embeddings) of a legal book.
3. 📖 Contextual Retrieval: Relevant sections of the legal text are retrieved based on semantic similarity to the user's query.
4. 🧠 AI-Powered Response: The original query and retrieved context are passed to GPT-Neo 1.3B, which generates a response grounded in legal context.

---

⚙️ Tech Stack

• 🧾 Text Embeddings: Converts legal content into high-dimensional vectors for semantic search.
• 📊 Vector Database: Efficient document storage and retrieval using FAISS, Pinecone, or similar tools.
• 🧠 GPT-Neo 1.3B: Open-source transformer model to generate legal insights.
• 🐍 Python: Core language for backend logic and system orchestration.

---

💼 Use Cases

• 🧑‍🎓 Law Student Support: Quickly understand legal concepts and references.
• 🧾 Legal Research Assistant: Automate lookup of relevant sections from legal texts.
• 🧠 General Legal Literacy: Make legal knowledge more accessible to non-lawyers.

---

⚠️ Disclaimer: This tool provides AI-generated responses for educational and informational purposes only. It is not a substitute for professional legal advice.

---

📁 Directory Structure

rag-system/
├── chunking/          # Document loading and chunking
├── embeddings/        # Embedding logic
├── vector_store/      # FAISS-based vector retrieval
├── prompts/           # Prompt templating
├── llm_interface/     # Local LLM loading (HuggingFace)
├── scripts/           # Main execution scripts
├── dataset/
│   └── raw/           # Place your .pdf files here
├── requirements.txt
└── README.html
    

---

🚀 Getting Started

1. Install Dependencies

pip install -r requirements.txt

2. Place Documents

Put your .pdf files into:

data/raw/

3. Run the Pipeline

python scripts/run_rag.py

---

📌 Components

• Corpus Loader: Loads and chunks PDFs.
• Embeddings: Uses HuggingFace models like all-MiniLM-L6-V2.
• Vector Store: FAISS-powered similarity search.
• Prompting: Dynamic prompt construction from retrieved docs.
• LLM Inference: Runs a local model (e.g., EleutherAI/gpt-neo-1.3B) using Transformers pipeline.

---

🛠 Customize

You can replace:

• PDF loader with web or CSV input.
• FAISS with Chroma or other vector stores.
• LLM with OpenAI, Claude, LLaMA, etc.

---

📜 License

This project is open-source and available under the MIT License.

---

👨‍💻 Author

Mohsin Raza