Capstone· 45 min

BookingAgent — Hotel Reservation Chatbot

What you'll learn
  • ▸Run a LangChain agent with real tool calling and SQLite persistence
  • ▸Test four different tool functions through natural language
  • ▸See how the agent loop decides which tool to invoke

The BookingAgent is a real-world LangChain agent that manages hotel reservations through natural language. It uses SQLite for persistence and exposes four tools the model can call: check_availability, booking, cancel_booking, and enquiry_booking.

Architecture

  • LangChain agent loop — decides which tool to call based on user input
  • SQLite database (data/bookings.db) — stores rooms and reservations
  • Four tool functions — each wraps a SQL query
  • CLI chat interface — interactive back-and-forth with the agent
  • Bonus: Streamlit web UI for a visual interface

Setup

From inside the BookingAgent/ folder:

$ cd BookingAgent
$ python3 -m venv .venv && source .venv/bin/activate
$ pip install -e ".[openai]"
Choose your provider
Install openai, anthropic, or google extras: pip install -e ".[anthropic]" or pip install -e ".[google]"

Create a .env file:

.env
OPENAI_API_KEY=sk-...
BOOKING_AGENT_MODEL_PROVIDER=openai
BOOKING_AGENT_MODEL_NAME=gpt-4o-mini
BOOKING_AGENT_TEMPERATURE=0.1
BOOKING_AGENT_SHOW_TOOL_CALLS=true

Run the CLI agent

$ booking-agent

Example prompts to try

  • "Check availability for a deluxe room from 2026-04-10 to 2026-04-12 for 2 rooms"
  • "Book 1 standard room for John Doe from 2026-05-01 to 2026-05-04, email john@example.com"
  • "Cancel booking BK-ABC12345"
  • "Check the status of booking BK-ABC12345"

Available tools

  • check_availability — queries room availability by date range and type
  • booking — reserves rooms with guest details and returns a booking ID
  • cancel_booking — revokes a reservation by booking ID
  • enquiry_booking — looks up booking status by ID

Bonus: Streamlit UI

$ streamlit run src/booking_agent/streamlit_app.py

Opens a web chat interface where you can type the same prompts and see tool calls visualised in real time.

What this project teaches
Agent loop, tool routing, database persistence, and multi-provider support — all patterns from Sessions 5 and 6 applied in a production-shaped project.