AnasAli-AI/Egyptian-Telecom-Chatbot

Egyptian-Telecom-AI-Chatbot An LLM-powered bilingual chatbot that supports Egyptian Arabic (عامية مصرية) and English, designed to assist customers of all major telecom companies in Egypt. # Egyptian Telecom RAG 📱 A Retrieval-Augmented Generation (RAG) chatbot and assistant built to answer questions about Vodafone Egypt plans, services, and support. This project scrapes the latest data, processes it into semantic chunks, and uses a hybrid retrieval pipeline (Vector Search + BM25 + Reranking) combined with Groq LLM to generate highly accurate and context-aware responses. ## 🚀 Setup & Installation ### 1. Create a Virtual Environment It is recommended to use an isolated Python environment. You can use standard `venv` or `uv`: **Using `venv`:** ```bash python -m venv .venv # On Windows: .venv\Scripts\activate # On Mac/Linux: source .venv/bin/activate ``` ### 2. Install Dependencies If you are using standard `pip`: ```bash pip install -r requirements.txt ``` *(Note: If you use `uv`, you can quickly install everything using `uv sync` since a `uv.lock` file is present).* ### 3. Prepare Environment Variables (`.env`) You need API keys for the LLM models to generate answers and process vision tasks. Rename/copy the example environment file and fill in your keys: ```bash cp .env.example .env ``` Make sure to edit the `.env` file to add your keys: - **`GROQ_API_KEY`**: Get it from Groq Console - **`GEMINI_API_KEY`**: Get it from Google AI Studio --- ## 🗄️ Preparing the Database (Data Pipeline) Before you run the chatbot, you must populate the vector database with the scraped data. Run the following scripts in order: ### Step 1: Run Scrapers Scrape Vodafone's website (FAQs, plans, etc.) to get the raw data: ```bash python -m src.services.run_scrapers # if you need to update the data ``` ### Step 2: Clean the Data Clean and format the raw scraped data, format plans, and remove duplicates: ```bash python -m src.helpers.cleaner ``` ### Step 3: Chunk the Data Break the cleaned text into semantic passage chunks optimized for vector search: ```bash python -m src.services.chunker ``` ### Step 4: Create Embeddings Generate vector embeddings and save t …