LICHG4E/Narratives-around-the-Tunisian-Revolution

# Tunisian Revolution RAG System A question-answering system about the Tunisian Revolution (2010-2011) that supports Arabic, French, and English. The system uses RAG (Retrieval-Augmented Generation) to provide accurate answers based on historical documents. ## Project Overview This project implements a complete RAG pipeline that: 1. Loads and processes 22 documents about the Tunisian Revolution from multiple sources (Wikipedia, news articles) 2. Converts them into searchable vectors using multilingual embeddings 3. Uses **LM Studio with the Saka-14B model** to generate intelligent answers 4. Supports questions in Arabic, French, and English **Note:** The `screenshots/` folder contains testing results and the prompts used to develop this project. ## Quick Start Guide ### Prerequisites - Python 3.10+ - LM Studio (with Saka-14B model loaded) ### Step 1: Install Dependencies ```bash cd tunisian_rag pip install -r requirements.txt ``` ### Step 2: Start LM Studio 1. Open LM Studio 2. Load the Saka-14B model 3. Start the local server (default port: 1234) ### Step 3: Build the Database This creates the search index from all documents: ```bash python scripts/build_vector_db.py ``` Output: ``` [1/5] Loading documents... ✓ Loaded: 22 documents [2/5] Preprocessing documents... ✓ Cleaned text [3/5] Chunking documents... ✓ Created 368 chunks [4/5] Generating embeddings... ✓ Generated embeddings [5/5] Storing in vector database... ✓ Stored in ChromaDB BUILD COMPLETE ``` ### Step 4: Chat with the System ```bash python chat.py ``` Example conversation: ``` You: Who was Mohamed Bouazizi? Bot: Mohamed Bouazizi was a street vendor in Sidi Bouzid, Tunisia... You: متى بدأت الثورة؟ Bot: بدأت الثورة التونسية في 17 ديسمبر 2010... You: Où la révolution a-t-elle commencé? Bot: La révolution a commencé à Sidi Bouzid... ``` ## System Architecture ``` tunisian_rag/ ├── chat.py # Main chat interface ├── config/ │ └── config.yaml # LM Studio & model s …