Vega: AI Voice Agent

Vega is an AI-powered voice agent built with FastAPI and Python.
It uses Google Gemini API for natural language responses, Tavily API for web search, and OpenWeather API for real-time weather updates.

🚀 Features

• 🎙️ Conversational Persona: Detective Vega – a mysterious investigator with short, clever responses.
• 🔎 Web Search: Real-time Tavily integration with links to top results.
• ☁️ Weather Info: Current and forecast data from OpenWeather API.
• 🔊 Voice Interaction: Record audio directly from the web UI.
• 🔐 Session-based API Keys: Users can enter their own API keys for Gemini, Tavily, and OpenWeather.
• ⚡ Streaming Responses: Real-time LLM answers streamed back to frontend.

---

🛠️ Tech Stack

• Backend: FastAPI (Python)
• Frontend: HTML, JavaScript
• LLM: Google Gemini API
• Search: Tavily API
• Weather: OpenWeather API
• Deployment: Render

---

📂 Project Folder Structure

## 📂 Project Folder Structure

Aiagent/
├── main.py                # FastAPI entry point
├── config.py              # Configuration (API keys, settings)
├── requirements.txt       # Python dependencies
├── README.md              # Documentation
│
├── routes/                # API routes
│   ├── agent_chat.py      # Handles chat with LLM
│   └── transcriber.py     # Handles transcription
│
├── services/              # Core service modules
│   ├── llm_service.py     # Gemini LLM persona (Detective Vega)
│   ├── search_service.py  # Tavily web search integration
│   ├── weather_service.py # Weather API integration
│   └── tts_service.py     # Text-to-Speech streaming
│
├── static/                # Frontend files
│   ├── audio-processor.js # Audio processing for browser
│   ├── script.js          # Handles recording + WebSocket
│   └── style.css          # Optional styling
│
├── templates/             # (Optional) Jinja2 templates
│   └── index.html         # Web UI for user interaction
│
├── uploads/               # Stores uploaded files (if any)
│
├── utils/                 # Helper utilities
│
├── Agent/Output/          # Stores recorded audio locally
│   └── recorded_audio.webm
│
└── .env                   # Environment variables (ignored by Git)
     # (Optional) For persistent chat history storage

1️⃣ Clone the repository

git clone https://github.com/riya941/Aiagent.git cd Aiagent

2️⃣ Create and activate virtual environment

python -m venv venv

Windows

venv\Scripts\activate

Mac/Linux

source venv/bin/activate

3️⃣ Install dependencies

pip install -r requirements.txt

4️⃣ 🔑 API Keys Setup:

By default, this project requires you to enter API keys per session inside the code or when prompted.
This is useful for testing, but not secure for production.

Option 1: Enter keys per session

• Open config.py and paste your keys directly: -GEMINI_API_KEY = "your-gemini-key" -MURF_API_KEY = "your-murf-key" -WEATHER_API_KEY = "your-weather-key" -TAVILY_API_KEY = "your-tavily-key"

Option 2: Use environment variables

Create a .env file in the project root:

-GEMINI_API_KEY=your-gemini-key -MURF_API_KEY=your-murf-key -WEATHER_API_KEY=your-weather-key -TAVILY_API_KEY=your-tavily-key

These will be automatically loaded by config.py if present.

5️⃣ Run the FastAPI server

uvicorn main:app --reload

6️⃣ Open the application in browser

http://127.0.0.1:8000

🎤 Usage Examples

🌐 Web Search

• "Latest Tesla stock price"
• "Search news about AI startups"

☁️ Weather

• "Weather in New York"
• "Forecast for Tokyo tomorrow"

🧠 General Knowledge

• "Who is the president of India?"
• "What is the national animal of India?"

---

🌍 Deployment (Render)

1. Push your project to GitHub.
2. Connect repo to Render.
3. Render automatically builds & deploys.

ℹ️ Note:

• If redeployment fails, old version keeps running ✅
• URL remains the same ✅