nabil-stem/darija-callcenter-voice-ai

# Darija Call Center Voice AI Production-oriented Moroccan Darija voice support assistant for call center workflows. This project combines sentiment analysis, LLM response generation, speech-to-text, and text-to-speech behind a FastAPI backend and lightweight web console. ## Why This Project Darija customer support automation is often blocked by low-resource dialect tooling. This repository provides a practical, end-to-end pipeline to prototype and deploy real-time Darija voice interactions for e-commerce and service support use cases. ## Core Features - Darija customer text and audio input handling - Sentiment-aware support reply generation - Optional speech synthesis for spoken agent responses - FastAPI endpoints for integration with external apps - Browser-based console for quick testing and demos - Dataset build and LoRA fine-tuning utilities ## Tech Stack - Python + FastAPI - Transformer-based LLM workflow - Speech transcription and voice generation pipeline - Static frontend (`web/`) for operator testing ## Quick Start ```powershell cd powershell -ExecutionPolicy Bypass -File .\setup_callcenter_env.ps1 .\.venv_callcenter\Scripts\python.exe -m uvicorn callcenter_api:app --host 127.0.0.1 --port 8000 ``` In a second terminal: ```powershell cd \web python -m http.server 5500 ``` Then open `127.0.0.1` and set API URL to `127.0.0.1`. ## API Endpoints - `GET /health` - `POST /reply-text` - `POST /reply-voice` ## Repository Notes - Runtime audio, model checkpoints, and generated artifacts are excluded from git by default. - Keep sensitive credentials and production keys out of this repository. ## Author **Nabil Essaadi** ## SEO Keywords Darija AI, Moroccan Arabic NLP, call center AI, voice agent, customer support automation, FastAPI voice API, speech-to-text Darija, text-to-speech Darija