🧓 Silverline – A Voice of Hope for the Elderly

Silverline is an AI-powered voice-to-voice assistant designed to help elderly individuals in moments of need—whether emotional, medical, environmental, or related to daily living.

---

💡 The Problem

In regions like Pakistan, many elderly individuals—and even younger people in remote areas—lack access to or understanding of modern technology. In emergencies, app-based solutions and online tools are often out of reach. Emergency services can be unreliable or delayed. For these communities, a simple phone call can mean the difference between help and helplessness.

---

✨ The Silverline Solution

Silverline makes accessing support as simple as dialing a number—no internet, smartphone, or technical skills required.

✅ No apps
✅ No downloads
✅ No typing
✅ Just speak

---

〰 Agentic Architecture Diagram

---

🗣️ What Makes Silverline Special?

🎙️ Real-Time Voice AI (Powered by OpenAI)

• Uses OpenAI’s real-time API for natural, two-way voice conversations.
• Preserves tone, sentiment, and emotional nuance—no loss through text conversion.
• Responds in a soothing and empathetic voice with 7 voice options (male and female).
• Built from scratch using custom WebSocket handling—no external wrappers like Vapi used.

---

🧩 Enhanced by AgentPro – Real-Time Knowledge Retrieval

Integrates AgentPro—a modular agent framework—to enable:

🔍 Real-time internet search, so the bot can assist with dynamic, location-specific queries such as:

• 📍 Nearby hospitals or pharmacies
• 🌤 Weather and traffic updates
• 📘 Public service information or general queries

By using AgentPro’s tool-calling capabilities, Silverline overcomes model knowledge cutoffs and provides live, up-to-date information from the web.

---

📞 Additional Call Handling Features

• Spam Prevention:
  Silverline employs robust spam call detection mechanisms to ensure that the system does not entertain unsolicited or harmful spam calls.

• Automatic Query Detection:
  The system automatically detects if a caller's query requires an external internet search. When needed, it seamlessly invokes AgentPro to retrieve relevant, up-to-date data.

• Call Classification:
  Silverline classifies calls based on conversational context. This classification supports tailored responses and efficient routing of requests, ensuring that each call is handled appropriately, whether it's a general inquiry, emergency, or alarm-triggered situation.

---

---

📊 Dashboard & Call Analytics

Track all interactions in real time:

Caller Number	Twilio Number	Type	Duration	Time	Spam Status	Reason

• Detects and blocks spam calls.
• Categorizes each interaction by type and intent.

---

🛠️ Backend Logging

Why are we showing logs?
Because everything powering Silverline is custom-engineered—especially the voice interface. Unlike plug-and-play wrappers (like Vapi), Silverline handles OpenAI's real-time streaming API directly via WebSockets.

The log viewer is designed to show only AI Backend logs, with the drop-down default set to "AI Backend" because the entire logic of real-time communication is handled there. Although other backend options are available as drop-down choices, they are not active, as the focus is demonstrating the AI handling of events.

These logs showcase:

• Event-driven handling of user voice input
• Orchestration of WebSocket messages
• Prompting, streaming, and robust edge-case recovery
• Asynchronous performance with detailed logging

It’s a demonstration of the benefits of building a deeply integrated voice intelligence system from the ground up.

---

🧬 Tech Stack

• Voice AI Model: OpenAI GPT (Real-Time API)
• Telephony: Twilio
• Agent Framework: AgentPro
• Frontend: Next.js
• Backend: FastAPI (Async)
• Logging: Custom real-time log capture
• Infrastructure: Docker + Nginx + Async Workers

---

🧪 Run It Locally

🔻 Clone the Repository

git clone https://github.com/mubashirsidiki/Traversaal-x-Optimized-AI-Hackathon.git
cd Traversaal-x-Optimized-AI-Hackathon

---

🛠️ Set Up Environment Variables

📂 In /backend/.env

# OpenAI API Key:
OPENAI_API_KEY=

# OpenAI Realtime Model:
OPENAI_REALTIME_MODEL=

# Backend URL:
backend_url=

# Classifier Model:
CLASSIFIER_MODEL=

# MongoDB Connection:
DATABASE_URL=

# Traversaal Ares API Key:
TRAVERSAAL_ARES_API_KEY=

📂 In /frontend/.env

# API Configuration
NEXT_PUBLIC_API_PREFIX=/ai/api
NEXT_PUBLIC_API_TIMEOUT=90000

# URLs
NEXT_PUBLIC_BASE_URL=http://localhost:8000
NEXT_PUBLIC_DIALPAD_URL=http://localhost:3001/dialer
NEXT_PUBLIC_LOG_VIEWER_URL=http://localhost:3002/gen-log

📂 In /dialpad/.env

# Twilio Configuration
TWILIO_ACCOUNT_SID=
TWILIO_AUTH_TOKEN=
VALID_CALLER_ID=
TWILIO_API_KEY=
TWILIO_API_SECRET=
TWILIO_TWIML_APP_SID=

# Server Configuration
PORT=3001

📂 In /general-logging/.env

# API Endpoints
AI_BACKEND_API=http://localhost:8000/ai/server/send-logs
BACKEND_API=
FRONTEND_API=

---

🐳 Run with Docker (Recommended)

docker-compose up --build -d

To stop:

docker-compose down

---

🧰 Run Without Docker

1. Backend

cd backend
poetry install
poetry run python main.py

2. Frontend

cd ../frontend
npm install
npm run build
npm run serve

3. Dialpad

cd ../dialpad
npm install
node server/server.js

4. General Logging

cd ../general-logging
npm install
node server.js

---

🌐 Local Access URLs

After successful configuration and running the project locally, you can access:

• Frontend at: http://localhost:3000
• Backend at: http://localhost:8000
• Dialer at: http://localhost:3001
• Log Viewer at: http://localhost:3002

---

🌍 Vision

“In a world driven by screens, Silverline gives hope through a voice.”

Silverline isn’t just a tech demo—it’s a mission to make AI accessible, human, and life-saving for those who need it most.

---