Bandoh/constitution-rag-ai-system

🇬🇭 Semantic search and Q&A system for Ghana's Constitution. Built with Ollama LLMs, LlamaIndex RAG pipeline, and Qdrant vector database. Runs entirely locally with natural language queries and context-aware responses. # Ghana Constitution RAG System A Retrieval-Augmented Generation (RAG) system for querying Ghana's Constitution using local LLMs and vector search. ## Overview This project implements a conversational AI system that answers questions about Ghana's Constitution. It uses: - **Qdrant** for vector storage and similarity search - **LlamaIndex** for document processing and retrieval - **Ollama** for local LLM inference - **HuggingFace embeddings** for document vectorization ## Features - 📚 Semantic search through Ghana's Constitution - 🤖 Natural language responses powered by local LLMs - 🔍 Context-aware retrieval with top-k similarity matching - 💬 Conversational interface with legal expertise - 🔒 Runs entirely locally - no external API calls ## Prerequisites - Python 3.8+ - Docker (for Qdrant) - Ollama installed locally - Intel XPU support (or modify to use CPU/GPU) ## Installation ### 1. Clone the repository ```bash git clone github.com cd constitution-rag-ai-system ``` ### 2. Install Python dependencies ```bash pip install -r requirements.txt ``` ### 3. Install and Setup Ollama **For Linux:** ```bash curl -fsSL ollama.com | sh ``` **For macOS:** ```bash brew install ollama ``` **For Windows:** Download from ollama.com ### 4. Start Ollama Server ```bash ollama serve ``` Keep this terminal running. Open a new terminal for the next steps. ### 5. Pull an Ollama Model Choose one of these models (or any other from ollama.com): **Recommended models:** ```bash # Fast and efficient (3.8GB) ollama pull llama3.2 # More capable (4.7GB) ollama pull mistral # Larger, more accurate (7.4GB) ollama pull llama3.1:8b # Powerful option (26GB) ollama pull llama3.1:70b ``` For this example, we'll use `llama3.2`:github.com ```bash ollama pull llama3.2 ``` ### 6. Start Qdrant Vector Database ```bash docker run -p 6333:6333 -p 63 …