RamiBenAmor/Tunisian-AI-CSR-hackathon

# 🗣️ AI-Powered Voice Chatbot for Tunisie Telecom An intelligent, multilingual voice chatbot designed to improve access to Tunisie Telecom services — especially for Tunisian Arabic speakers. Built with a focus on accessibility, language understanding, and real-time interaction. ## 🏆 Hackathon Recognition 🥈 **2nd Place** — AI-RSE/ESG Hackathon, CSR Power Forum 2025 📍 Organized by Forum international RSE & Tunisie Telecom 👥 Only all-student team in the Top 3 --- ## 🚀 Overview This project demonstrates the power of conversational AI in local language contexts. The chatbot supports both **French** and **Tunisian dialect**, offering voice and text interfaces for streamlined interaction with telecom services. --- ## 🔧 Features - **🗣 Multilingual Support** Conversational support in **Tunisian Arabic** 🇹🇳 and **French** 🇫🇷 - **🔍 Automatic Language Detection** Dynamically adapts to the user's spoken or written language - **🎯 Smart Response Generation** Integrates the **Gemini API** for context-aware, intelligent replies - **🧠 Dynamic Knowledge Routing** Queries are classified and routed to the appropriate knowledge source - **🎤 Voice Interaction Pipeline** Full **Speech-to-Text** and **Text-to-Speech** integration - **🌐 Real-time Demo Interface** Web demo powered by a lightweight **Flask** backend --- ## 🛠 Tech Stack - Python, Flask - Gemini API - SpeechRecognition, gTTS / TTS - SentenceTransformers, FAISS - Language detection with `langdetect` - Tunisian dialect processing