koushik444/speech-ai-backend

VoxFlow AI 🎙️⚡

A high-performance, real-time speech-to-speech AI platform optimized for Hugging Face Spaces (Backend) and Vercel (Frontend).

🚀 Features

• —Real-time Voice Chat: Low-latency WebSocket-based audio streaming.
• —Cloud Powered: Optimized with Groq (STT/LLM) and Edge-TTS for lightning-fast, free-tier performance.
• —Monolith Architecture: Single container running Gateway, STT, LLM, and TTS services.

🛠 Deployment Guide

Backend (Hugging Face)

1. 1.Create a new Docker Space on Hugging Face.
2. 2.Upload this repository.
3. 3.Add the following Secrets in Settings:
4. 4.GROQ_API_KEY: Your key from Groq Cloud.
5. 5.REDIS_URL: Your REST URL from Upstash Redis.
6. 6.Add the following Variables in Settings:
7. 7.STT_PROVIDER: groq
8. 8.LLM_PROVIDER: groq
9. 9.PROVIDER: edge-tts

Frontend (Vercel)

1. 1.Import this repository to Vercel.
2. 2.Set the Root Directory to frontend.
3. 3.Add Environment Variables:
4. 4.NEXT_PUBLIC_API_URL: Your HF Space URL (https).
5. 5.NEXT_PUBLIC_WS_URL: Your HF Space URL (wss).

📂 Project Structure

• —`frontend/`: Next.js 14 web application.
• —`backend/`: Python microservices (Gateway, STT, LLM, TTS) consolidated for monolith deployment.
• —`Dockerfile`: Root Docker configuration for Hugging Face Spaces.
• —`supervisord.conf`: Process manager configuration for the backend monolith.

This Space runs a FastAPI monolith managed by supervisord.