Waypoint is a fast, accurate, and responsive LLM Travel Assistant designed for seamless map-based data visualization.

It’s built using LangChain, LangGraph, and TypeScript, with a clean and efficient UI powered by Next.js, React, and Tailwind CSS. The backend leverages Google Cloud services to provide robust multi-tool capabilities and rapid response times. The application processes user queries in near real-time, allowing for fluid interactions and instantaneous map updates.

Pipeline

Waypoint’s query-answer pipeline follows a few simple steps. We first load the LLM with a few custom-built tools such as a weather and pinpoint tool. Next we give it the user query. The LLM maintains a chat history to provide contextually relevant responses.

Next, the LLM is given the choice to use a tool or respond with a plain text response via a conditional edge. Depending on the tool and query, we can also have additional processes that give a better user experience.

Demo

Other tools

Below is an example of the generative UI capabilities. The LLM uses a WeatherTool to give the user a downloadable Weather Card.