About Lighthouse Bot
AI-Powered Maritime Data Analysis Platform
What is Lighthouse Bot?
Lighthouse Bot is a data analysis application that uses natural language queries to analyze maritime ferry data using agentic Retrieval-Augmented Generation (RAG). The platform allows users to ask questions about ferry operations, routes, passenger traffic, and performance metrics in plain English, and receive detailed analytical responses backed by real data.
This innovative platform bridges the gap between complex maritime data and accessible insights, making data analysis available to both technical and non-technical users in the maritime industry.
How to Use This Platform
🔑 API Key Requirement
To use this application, you need an OpenRouter API key. This key enables access to various Large Language Models (LLMs) for processing your queries.
Note: The platform is designed for querying and analysis only. Advanced features like model evaluation are available through the source code.
📊 Getting Started
- Select a language model from the dropdown menu
- Enter your question about ferry data (e.g., "What is the average speed of ferry Jupiter?")
- View comprehensive responses including SQL queries and analysis
- Access your query history to track previous analyses
📊 Ferry Database Overview
🚢 Ferry Fleet Data
🔗 View Source DataOur database contains real-world operational data from Färjerederiet's ferry fleet, covering the period from March 2023 to February 2024.
🚢 Available Ferries with Trip Data (5 Vessels)
Note: Additional ferry specifications are available for 6 more vessels (Skidbladner, Marie, Capella, Linda, Sedna, Ebba Brahe) but without trip data.
📈 Available Data Categories
💡 Example Questions You Can Ask
Performance Analysis:
- • "What is the average fuel consumption of ferry Jupiter?"
- • "Which ferry has the best fuel efficiency?"
- • "Compare the average speed of all ferries"
- • "Show me the longest trips by distance"
Operations & Traffic:
- • "What are the busiest routes by passenger volume?"
- • "How many extra trips were made in 2023?"
- • "Which ferry carries the most vehicles on average?"
- • "Show passenger patterns by month"
🗃️ Key Database Fields
Trip Information:
time_departure
trip_type
tailored_trip
start_time_outbound
end_time_outboundPerformance Data:
distance_outbound_nm
fuelcons_outbound_l
distance_inbound_nm
fuelcons_inbound_lLoad & Capacity:
passenger_car_equivalent_outbound
passenger_car_equivalent_inbound
vehicles_left_at_terminal_outbound
vehicles_left_at_terminal_inboundLearn More
📚 Full Documentation & Source Code
For detailed setup instructions, evaluation features, and technical documentation:
🔗 GitHub Repository - vildly/lighthouse-sensor-bot🗃️ Raw Ferry Data Source
The ferry operational data used in this platform is part of the Hack-A-Fleet v2.0 dataset from RISE Maritime:
🔗 RISE Maritime - Hack-A-Fleet Dataset (ferry_trips_data.csv)🎓 Academic Research
This platform was developed as part of a bachelor's degree thesis evaluating LLMs for maritime data analysis:
📄 "Lighthouse Bot: A Platform For Evaluating LLMs For Agentic Maritime Data Analysis"Contributors
Oxana Sachenkova
Supervisor & Contributor
Dongzhu Tan
Co-Developer & Researcher
Melker Andreasson
Co-Developer & Researcher
This project was developed as part of a Computer Science bachelor's degree program, focusing on evaluating Large Language Models for maritime data analysis applications. Oxana Sachenkova served as the academic supervisor, guiding the research and contributing to the application development.
Technical Overview
Frontend
Next.js interface for submitting queries and viewing results
Backend
Flask server processing queries with LLM agents
Database
PostgreSQL storing query results and evaluations