AhmedBensalah8603/Tunisian_STT

A robust Automatic Speech Recognition (ASR) engine for the Tunisian dialect, featuring code-switching support, a FastAPI-based and deployment configurations for Docker, Azure, and Kubernetes. # 🇹🇳 Tunisian STT API Service This repository contains a production-ready **FastAPI** service for **Tunisian Arabic Speech-to-Text (ASR)**. It wraps a specialized Vosk/Kaldi model (`STT_Tun_Model`) into a REST API and provides complete configuration for deployment on **Azure**. ## 🚀 Features * **FastAPI Backend**: High-performance, asynchronous REST API. * **Tunisian Dialect Support**: Optimized for Tunisian Arabic with code-switching (French/English). * **Dockerized**: Complete `Dockerfile` for consistent environments. * **Azure Ready**: Scripts and documentation for deploying to Azure Container Instances (ACI). * **Streaming Support**: Real-time transcription capabilities. ## 📂 Project Structure ``` Tunisian_STT/ ├── app.py # FastAPI application ├── Dockerfile # Docker configuration ├── requirements.txt # Python dependencies ├── deploy_to_azure.sh # Automated Azure deployment script ├── testing.py # CLI script for testing the model └── test_api.py # Script to test the deployed API ``` ## 🛠️ Local Setup ### Prerequisites * Python 3.10+ * Docker (optional, for containerization) ### 1. Install Dependencies ```bash pip install -r requirements.txt ``` ### 2. Prepare the Model You can download the model directly using `curl`: ```bash sudo apt-get install curl curl -L huggingface.co \ --output STT_Tun_Model.zip ``` Then extract it and organize the directory structure: ```bash unzip STT_Tun_Model.zip mkdir -p model mv STT_Tun_Model model/ ``` ### 3. Run Locally ```bash uvicorn app:app --host 0.0.0.0 --port 8000 ``` Access the API documentation at `localhost`. --- ## ☁️ Azure Deployment This project includes a fully automated script to deploy the service to **Azure Container Instances (ACI)**. ### Prerequisites * **Azure CLI**: Install Azure CLI * **Azure Account**: You need an active subscription. …