Flash: AI-Powered Voice Assistant

Overview

Flash is an AI-powered voice assistant built on a FastAPI backend that integrates multiple cloud-based services. The project combines:

• Twilio for call handling and webhook integration,
• Google Cloud Speech-to-Text for transcribing user recordings,
• Groq's AI API for generating conversational responses,
• Google Cloud Platform (GCP) for hosting the FastAPI service and storing transcripts/recordings.

This repository contains the complete implementation of the basic flows, as demonstrated in the video demo.

Features

• FastAPI Backend: A robust backend built with FastAPI for handling incoming Twilio webhook requests.
• Twilio Integration: Seamless call handling with Twilio, including webhook processing for incoming calls and recordings.
• Google Cloud Storage & Transcription: Recordings are fetched from Twilio and processed using Google Speech-to-Text, allowing voice-to-text conversion.
• Groq AI Assistant: Uses Groq's API to generate conversational replies based on the transcribed text.
• GCP Hosting: The FastAPI app is deployed on Google Cloud, leveraging free tiers (with some performance trade-offs).
• Unified Flows: Demonstrates the complete integration—from incoming voice call to AI-generated response—through a cohesive flow.

Setup & Installation

Prerequisites

• Python 3.9 or newer
• Twilio account with valid credentials
• Google Cloud Platform account with Speech-to-Text API enabled
• Groq API key for AI responses

Environment Variables

Create a .env file with the following:

TWILIO_ACCOUNT_SID=your_account_sid
TWILIO_AUTH_TOKEN=your_auth_token
GROQ_API_KEY=your_groq_api_key
GOOGLE_APPLICATION_CREDENTIALS=path_to_credentials.json
GCP_BUCKET_NAME=your_bucket_name
WEBSOCKET_URL=your_ngrok_websocket_url