✈️ Agentic AI Travel Planner

A Multi-Agent, Tool-Augmented AI Travel Concierge — powered by LangGraph, Groq, and Real-Time APIs.

Solving the paradox of choice in travel planning with autonomous AI agents that research, calculate, and plan — so you don't have to.

📸 Visual Preview

🎬 The AI Concierge generating a 5-day Paris itinerary with live weather, map pins, and PDF export.

🏗️ System Architecture

graph TB
    subgraph "🖥️ Frontend — Flask :5000"
        A[🌐 Landing Page<br>GSAP + Tailwind + Glassmorphism] --> B[💬 Chat Interface<br>Markdown + Typing Indicator]
        B --> C[🗺️ Leaflet.js Map<br>Auto Marker Plotting]
        B --> D[📄 PDF Export<br>xhtml2pdf Branded Docs]
        A --> E[⚙️ Preferences Modal<br>SQLite Persistence]
    end

    subgraph "⚡ Backend — FastAPI :8000"
        F[🔀 API Router] --> G[🧠 LangGraph Agent<br>ReAct Loop]
        F --> H[📦 Preference CRUD<br>SQLAlchemy ORM]
        G --> I[🛠️ Tool Orchestrator]
    end

    subgraph "🛠️ Agent Tools"
        I --> J[🌤️ OpenWeatherMap<br>Current + 5-Day Forecast]
        I --> K[📍 Google Places<br>Search + Reviews + Coords]
        I --> L[💱 Frankfurter API<br>Real-Time Exchange Rates]
        I --> M[🧮 Calculator<br>Budget + Tax + Tips]
    end

    subgraph "💾 Persistence Layer"
        N[(🗄️ SQLite Database)]
        H --> N
        G -->|Inject Preferences| N
    end

    subgraph "🤖 LLM Provider"
        O[⚡ Groq Cloud<br>Llama 3.3-70B Versatile]
    end

    B -->|POST /query| F
    E -->|POST /preferences| F
    D -->|POST /export_pdf| F
    G --> O
    O --> G

    style A fill:#001F3F,stroke:#FFD700,color:#FFD700
    style B fill:#001F3F,stroke:#FFD700,color:#FFD700
    style G fill:#1a1a2e,stroke:#e94560,color:#FFD700
    style O fill:#0d0d0d,stroke:#e94560,color:#e94560

🎯 The Problem We Solve

The Paradox of Choice in Travel Planning.

Travelers today face 10+ tabs open simultaneously — hotel aggregators, weather sites, currency converters, review platforms, map tools. By the time they've gathered the data, decision fatigue has set in.

Our solution: A single conversational interface backed by autonomous AI agents that orchestrate real-time tool calls, synthesize data from multiple APIs, and deliver complete, data-driven travel plans — including weather forecasts, budgets in your currency, restaurant recommendations with ratings, and an interactive map with every location pinned.

This isn't a chatbot wrapper. It's a multi-tool agentic workflow built on LangGraph's ReAct pattern — where the AI decides which tools to call, in what order, and how to combine their outputs into a coherent plan.

✨ Key Features

🧠 Long-Term Memory Persistent user profiling powered by SQLite + SQLAlchemy. Your preferences (budget style, dietary needs, travel pace) are stored and dynamically injected into the LLM's system prompt — every plan is personalised.	🗺️ Interactive Mapping Leaflet.js with dark CARTO tiles and gold-themed markers. The AI response is parsed for coordinates in real-time, and pins are automatically dropped on the map. Navy-themed popups match the brand.
⚡ Lightning-Fast Inference Powered by Groq Cloud running Meta's Llama 3.3-70B Versatile. Sub-second token generation with the quality of a 70B parameter model. No GPU required on your end.	📊 Financial Dashboard Real-time currency conversion via the Frankfurter API + a precision calculator with tax, tip, and percentage tools. Complete cost breakdowns in your preferred currency.
🌤️ Live Weather Intelligence OpenWeatherMap integration for current conditions and 5-day forecasts. The agent automatically calls weather tools and factors conditions into itinerary recommendations.	📄 PDF Export One-click export of any AI-generated plan to a branded, print-ready PDF with Navy + Gold styling, structured tables, and a professional header/footer. Powered by `xhtml2pdf`.
📍 Smart Place Discovery Google Places API integration returning rich data: ratings, review counts, coordinates, types, and top review snippets. The agent uses both text search and detail endpoints.	🎨 Premium UI/UX Glassmorphism design with GSAP animations, particle canvas background, Tailwind CSS, responsive layout, and a floating CTA. Designed to feel like a luxury concierge service.

🔧 Tech Stack

Category	Technology
🤖 LLM
🧩 Agent Framework
⚡ Backend
🌐 Frontend
🗺️ Maps
📡 APIs
💾 Database
📄 PDF

🧬 Agentic Workflow — How It Works

This project uses LangGraph's ReAct (Reasoning + Acting) pattern to create a truly autonomous agent. Unlike simple prompt-chain applications, the agent:

Reasons about the user's request and decides which tools to call
Acts by executing tool calls (weather, places, currency, calculator)
Observes the results and decides whether to call more tools or respond
Iterates until it has gathered enough data for a comprehensive plan

User: "Plan 5 days in Paris for $2000"
  │
  ▼
┌──────────────────────────────────────────────────┐
│  🧠 LangGraph ReAct Agent (Llama 3.3-70B)       │
│                                                    │
│  Step 1: "I need weather for Paris"                │
│  → calls get_current_weather("Paris")              │
│  → calls get_weather_forecast("Paris", 5)          │
│                                                    │
│  Step 2: "I need hotels and restaurants"            │
│  → calls search_places("hotels", "Paris")          │
│  → calls search_places("restaurants", "Paris")     │
│                                                    │
│  Step 3: "I need to convert $2000 to EUR"           │
│  → calls convert_currency(2000, "USD", "EUR")      │
│                                                    │
│  Step 4: "Let me calculate per-day budget"          │
│  → calls calculate("divide", 1850, 5)              │
│                                                    │
│  Step 5: Synthesise all data into a plan            │
│  → Returns complete Markdown itinerary             │
└──────────────────────────────────────────────────┘
  │
  ▼
📋 Day-by-day itinerary with weather, budgets, map pins, & PDF export

Guard Rails & Error Resilience

System Prompt Guard: The system prompt is always injected as the first message, regardless of state history
Tool Failure Guard: Empty or None tool responses are replaced with descriptive error strings
Preference Injection: User preferences from SQLite are dynamically appended to the system prompt per session
Graceful Degradation: If a tool fails (API down, timeout), the agent acknowledges it and proceeds with available data

📁 Project Structure

📦 ai-travel-planner/
├── 🤖 agents/
│   └── agentic_workflow.py     # LangGraph ReAct agent (GraphBuilder)
├── 🛠️ tools/
│   ├── calculator_tool.py      # Budget arithmetic tools
│   ├── currency_conversion.py  # Currency exchange tool
│   ├── place_search.py         # Google Places search + details
│   └── weather_information.py  # Weather current + forecast tools
├── ⚙️ utils/
│   ├── calculator.py           # Calculator with float-safe arithmetic
│   ├── currency_converter.py   # Frankfurter API wrapper
│   ├── place_info.py           # Google Places API (enriched)
│   ├── weather_info.py         # OpenWeatherMap API wrapper
│   ├── pdf_generator.py        # Markdown → Branded PDF
│   ├── model_loader.py         # LLM provider factory
│   ├── config_loader.py        # YAML config reader
│   └── save_document.py        # Document I/O utilities
├── 🌐 flask_app/
│   ├── app.py                  # Flask routes (query, prefs, PDF export)
│   ├── templates/
│   │   └── index.html          # Glassmorphism UI + Leaflet map
│   └── static/
│       ├── css/style.css       # Navy + Gold design system
│       └── js/main.js          # GSAP animations + Leaflet + Chat
├── 📝 prompt_library/
│   └── prompt.py               # SYSTEM_PROMPT for the AI agent
├── ⚙️ config/
│   └── config.yaml             # LLM model configuration
├── 📊 models.py                # SQLAlchemy ORM + PreferenceManager
├── 🚀 main.py                  # FastAPI application (API server)
├── 🖥️ app.py                   # Streamlit interface (alternative)
├── 📋 requirements.txt
├── 🔒 .env.example
└── 📖 README.md

🚀 Installation & Setup

Prerequisites

Python 3.11+
A Groq API Key (free tier available)
An OpenWeatherMap API Key (free tier)
A Google Places API Key (optional)

Step 1: Clone the Repository

git clone https://github.com/MustafaKocamann/ai-travel-planner.git
cd ai-travel-planner

Step 2: Create & Activate Virtual Environment

python -m venv venv

# Windows
venv\Scripts\activate

# macOS / Linux
source venv/bin/activate

Step 3: Install Dependencies

pip install -r requirements.txt

Step 4: Configure Environment Variables

cp .env.example .env

Edit .env with your API keys:

GROQ_API_KEY=gsk_your_groq_api_key_here
OPENWEATHERMAP_API_KEY=your_openweathermap_key_here
GOOGLE_PLACES_API_KEY=your_google_places_key_here  # optional

Step 5: Launch the Application

Open two terminals and run:

# Terminal 1 — FastAPI Backend (port 8000)
uvicorn main:app --reload --port 8000

# Terminal 2 — Flask Frontend (port 5000)
python flask_app/app.py

Step 6: Open in Browser

🌐 http://localhost:5000

💡 Tip: Complex trip plans (multi-day, multi-city) may take 1–3 minutes as the agent calls multiple tools sequentially.

💬 Usage Examples

Query	What the Agent Does
`"Plan 5 days in Paris for $2000"`	Weather + Hotels + Restaurants + Currency conversion + Budget breakdown
`"What's the weather in Tokyo?"`	Calls OpenWeatherMap API for current conditions
`"Best restaurants in Rome"`	Calls Google Places API with ratings and coordinates
`"Convert 500 USD to EUR"`	Calls Frankfurter API for live exchange rates

🗺️ Future Roadmap

🔜 v2.1	Voice AI integration — plan trips via voice commands using Whisper
📱 v3.0	Mobile-first PWA with offline support and push notifications
🌍 v4.0	Multi-language support (TR, DE, FR, ES, JA) with auto-detection
🏨 v5.0	Direct booking integration — hotels, flights, and activities
👥 v6.0	Collaborative trip planning — share and co-edit plans in real-time
📈 v7.0	Price prediction engine — ML-powered fare & hotel cost forecasting

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Fork the repository
Create your feature branch (git checkout -b feature/amazing-feature)
Commit your changes (git commit -m 'Add amazing feature')
Push to the branch (git push origin feature/amazing-feature)
Open a Pull Request

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

💛 Built with passion by Mustafa Kocaman

If this project helped you, consider giving it a ⭐ — it means the world!

Name		Name	Last commit message	Last commit date
Latest commit History 10 Commits
agents		agents
config		config
exception		exception
flask_app		flask_app
logger		logger
prompt_library		prompt_library
tools		tools
utils		utils
.env.example		.env.example
.gitignore		.gitignore
PROJECT_BRIEF.md		PROJECT_BRIEF.md
README.md		README.md
app.py		app.py
main.py		main.py
models.py		models.py
requirements.txt		requirements.txt

Folders and files

Latest commit

History

Repository files navigation

✈️ Agentic AI Travel Planner

📸 Visual Preview

🏗️ System Architecture

🎯 The Problem We Solve

✨ Key Features

🧠 Long-Term Memory

🗺️ Interactive Mapping

⚡ Lightning-Fast Inference

📊 Financial Dashboard

🌤️ Live Weather Intelligence

📄 PDF Export

📍 Smart Place Discovery

🎨 Premium UI/UX

🔧 Tech Stack

🧬 Agentic Workflow — How It Works

Guard Rails & Error Resilience

📁 Project Structure

🚀 Installation & Setup

Prerequisites

Step 1: Clone the Repository

Step 2: Create & Activate Virtual Environment

Step 3: Install Dependencies

Step 4: Configure Environment Variables

Step 5: Launch the Application

Step 6: Open in Browser

💬 Usage Examples

🗺️ Future Roadmap

🤝 Contributing

📄 License

💛 Built with passion by Mustafa Kocaman

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages