Inspiration
Inspired by the inherent complexity of air travel, FlightRag aims to demystify global flight networks. We set out to build a tool that empowers both non-technical users and experienced data scientists to explore and understand these connections, enabling them to gain insights quickly and efficiently. Imagine learning how competitor airlines are doing or discovering unique knowledge to enhance your own services.
What it does
FlightRag is a platform that translates natural language questions into interactive map visualizations and Plotly graphs of flight networks.
How we built it
We leverage LangChain and LangGraph for intelligent planning and reasoning, and the Rewoo agentic framework for dynamic task execution with minimal token usage. Our GPU-powered graph database on Arangodb ensures rapid data analysis, enabling seamless exploration of massive datasets. We've also integrated a weather tool to provide real-time, historical, and forecasted weather data.
Challenges we ran into
A primary challenge was translating natural language queries into precise graph operations. We addressed this by refining our LangChain prompts and developing a robust task planning module within LangGraph. Optimizing performance for large datasets was another challenge, which we tackled by leveraging GPU acceleration and efficient graph traversal algorithms. Generating Structured output is crucial in accomplishing complex task.
Accomplishments that we're proud of
FlightRag democratizes access to flight network data. Both non-technical users and data scientists can ask questions and receive interactive map visualizations and Plotly graphs, without writing code. This empowers them to extract valuable insights, opening up new possibilities for research, strategic planning, and operational decision-making.
What we learned
Agentic framework matters in how a user query intent is parsed to plans and executed. Reasoning to understand user intent and developing the rights workflow is key.
What's next for FlightRag
We plan to enhance FlightRag with more advanced analytical features, including predictive modeling and real-time data integration. We believe social analytics can be further improved by incorporating passenger volume and sentiment data.
Log in or sign up for Devpost to join the conversation.