🚀 Flight AI Assistant

Table of Contents

1. What is this repo?
2. Why function‑calling tools are the secret sauce?
3. Core Architecture – A Quick Tour
4. New Feature Spotlight: Discount‑For‑Sales Tool ✨
5. Getting Started – Run It Locally in 5 Minutes
6. Extending the Assistant – Add Your Own Magic
7. Best‑Practices, Gotchas & Tips
8. License & Contributing

---

What is this repo?

Flight AI Assistant is a playful yet powerful chatbot that demonstrates how to blend large‑language‑model chat (OpenAI GPT‑4o‑mini, Anthropic Claude, Google Gemini) with function‑calling tools to fetch real‑time, reliable data.
Think of it as a virtual airline agent that never makes up ticket prices—because the prices live in a tiny, well‑structured Python dictionary (or a future live API).

TL;DR: Chat with a model, let it call Python functions for you, and get spot‑on answers.

---

Why function‑calling tools are the secret sauce?

✅ Benefit	🎯 What it means for you
Separation of concerns	The system prompt handles tone (courteous, one‑sentence answers). The tools handle the hard facts (prices, discounts).
Zero hallucinations	The model never guesses a price; it must call get_ticket_price or get_discounted_price.
Structured, JSON‑friendly data	Responses are machine‑readable, no fragile regex parsing required.
Easy extensibility	Add a new function → update its JSON schema → append to tools. No system‑prompt rewrites!
Auditability	Every tool call can be logged for compliance or debugging.
Security by design	The model never gets direct DB or network access—only the curated functions you expose.

---

Core Architecture – A Quick Tour

User → Gradio UI → chat() → OpenAI API
           ↘ (no tool)               ↘ (tool call)
             ↘                         ↘
        OpenAI response          handle_tool_call()
             ↘                         ↘
   UI displays answer   tool response → second OpenAI call → final answer

• system_message – sets the personality: short, courteous, always accurate.
• Tool definitions – JSON schemas (price_function, discount_function) describing the callable Python functions.
• chat() – orchestrates the conversation, decides when a tool call is needed, and stitches everything together.
• handle_tool_calls() – parses the model’s tool request, runs the real Python function, and builds the tool‑response payload.
• Gradio UI – a sleek web front‑end (gr.ChatInterface) for instant testing.

---

New Feature Spotlight: Discount‑For‑Sales Tool ✨

What’s new?

We’ve added a second tool that returns discounted ticket prices for special promotions (Black Friday, Christmas sales, etc.).

How it works

1. User asks “What’s the sale price for a Tokyo ticket?”
2. The model sees the discount_function schema, decides to call it.
3. handle_tool_calls() runs get_discounted_price(city) against the discount_prices dict.
4. The model receives the JSON result and crafts a short, friendly reply like:

“A Tokyo round‑trip is currently on sale for $999 🎉”

Tool definition (for nerds)

discount_function = {
    "name": "get_discounted_price",
    "description": "Get the discounted price of a return ticket to the destination city. Use this during sales periods (e.g., Black Friday, Christmas).",
    "parameters": {
        "type": "object",
        "properties": {
            "destination_city": {
                "type": "string",
                "description": "The city the customer wants to travel to."
            }
        },
        "required": ["destination_city"],
        "additionalProperties": False,
    },
}
tools += [{"type": "function", "function": discount_function}]

Now the assistant can smoothly switch between regular and discounted pricing without any extra prompt gymnastics.

---

Getting Started – Run It Locally in 5 Minutes

1. Clone the repo

   git clone https://github.com/AsutoshaNanda/Flight-AI-Assistant.git
   cd Flight-AI-Assistant

2. Create a virtual environment (optional but recommended)

   python -m venv venv && source venv/bin/activate   # macOS/Linux
   .\venv\Scripts\activate                          # Windows

3. Install dependencies

   pip install openai anthropic google-generativeai gradio python-dotenv beautifulsoup4

4. Add your API keys in a .env file at the project root:

   OPENAI_API_KEY=sk-...
   ANTHROPIC_API_KEY=sk-ant-...
   GOOGLE_API_KEY=AIzaSy...
   

5. Launch the notebook or script
   Notebook: jupyter lab → open Flight AI Assistant.ipynb → run all cells.
   Script: copy the notebook code into a app.py and run python app.py.

6. Open the Gradio UI (usually at http://127.0.0.1:7860) and start chatting!

---

Extending the Assistant – Add Your Own Magic

1. Write a Python function – e.g., def get_flight_status(flight_number): ...
2. Create a matching tool schema (name, description, JSON parameters).
3. Append it to the tools list.
4. Update handle_tool_calls() to route the new function.name to your Python function.

Tip: Keep the function pure (no side‑effects) and validate inputs (whitelist city names, regex flight numbers) for security.

---

Best‑Practices, Gotchas & Tips

🎯 Area	✅ Recommendation
Infinite loops	Limit max_tool_calls per turn (e.g., 2) to avoid recursion.
Schema mismatches	Test tool schemas in OpenAI Playground before coding.
User‑supplied data	Sanitize/whitelist all arguments inside your functions.
Deprecation warnings	Use type='messages' in gr.ChatInterface (as shown).
Rate limits	Implement exponential back‑off & exponential jitter for API calls.
Debugging	Print/log the raw tool_calls payload to see exactly what the model sent.
Fun factor	Sprinkle emojis, friendly language, and short sentences (the system prompt already enforces it!).

---

License & Contributing

This project is licensed under the Apache License 2.0 – see the LICENSE file for full details.

💡 Contributions welcome! Feel free to fork, add new tools (e.g., flight status, baggage allowance), improve the UI, or write tests. Open an issue or submit a pull request, and let’s make this little airline agent even smarter together.

---

Happy coding ✈️💸

Please contribute to this and hope this repo would be some help of yours.