Built entirely with natural language. Zero manual coding.
Vibe Coding is a new paradigm of software development where you describe what you want in plain English and AI builds it for you.
- ❌ No memorizing syntax
- ❌ No debugging semicolons for hours
- ✅ You are the architect — AI is the engineer
The barrier to building software is no longer technical skill. It is the clarity of your thinking.
This is a web-based AI Flight Search System built using a multi-agent architecture — powered by Google Gemini 2.5 Flash and the Duffel API.
You type something like:
"I want to fly from Montreal to Toronto next Friday under $300"
And the system:
- ✅ Understands your intent using AI
- ✅ Converts it into a structured search query automatically
- ✅ Queries live flight data from real airline databases
- ✅ Returns flights with prices, times & stops
- ✅ Displays everything in a clean two-panel web interface
No forms. No dropdowns. No manual filters. Just talk to it like a human.
User (Natural Language)
│
▼
┌─────────────────────────────────┐
│ Agent 1 — Orchestration │
│ Gemini 2.5 Flash (LLM) │
│ │
│ Perception → Intent Parsing │
│ → Structured JSON Output │
│ { origin, destination, │
│ departure_date, passengers, │
│ price_limit } │
└────────────────┬────────────────┘
│ Structured JSON
▼
┌─────────────────────────────────┐
│ Agent 2 — Flight Search │
│ │
│ Request Handler → API Caller │
│ → Raw Flight Results │
└────────────────┬────────────────┘
│ API Call
▼
┌────────────────┐
│ Duffel API │
│ (Live Flights) │
└────────────────┘
│ Results
▼
┌─────────────────────────────────┐
│ Web UI — Flask + Python │
│ Left: Flight Results Cards │
│ Right: AI Chat Interface │
└─────────────────────────────────┘
| Component | Technology |
|---|---|
| LLM (AI Brain) | Google Gemini 2.5 Flash |
| Flight Data | Duffel API |
| Backend | Python + Flask |
| AI Development Tool | Gemini CLI |
| Config Management | .env file |
- Python 3.10+
- A Gemini API Key
- A Duffel API Key
1. Clone the repository
git clone https://github.com/iamkarandeepsingh/Vibe-Coding-
cd Vibe-Coding-2. Install dependencies
pip install -r requirements.txt3. Set up your API keys
Create a .env file in the root directory:
GEMINI_API_KEY=your_gemini_api_key_here
DUFFEL_API_KEY=your_duffel_api_key_here4. Run the application
python app.py5. Open in your browser
http://localhost:5000
"I want to fly from Montreal to Toronto next Monday"
"Find me a flight from London to Paris this weekend under $200"
"I need a return flight from New York to LA, departing March 20"
Vibe-Coding-/
├── app.py # Main Flask application
├── agents/
│ ├── agent_1_intent_parser.py # Orchestration agent
│ └── agent_2_flight_search.py # Flight search agent
├── templates/ # HTML web interface
├── static/ # CSS and JS assets
├── requirements.txt # Python dependencies
└── .env # API keys (not committed)
- Gemini API — Free tier includes 1,000 requests/day via Google AI Studio
- Duffel API — Test mode provides sandbox flight data. Sign up at duffel.com
⚠️ Never commit your.envfile or share your API keys publicly.
This project is proof that in the AI era, anyone with a clear idea and the ability to communicate it can build production-level software.
"The builders of tomorrow won't be the ones who memorized the most syntax. They'll be the ones who asked the best questions." 🧠
MIT License — feel free to use, modify, and build on this project.
- Google Gemini — for the LLM powering the agents
- Duffel — for the real-time flight data API
- Gemini CLI — the vibe coding tool used to build this entire project