Ask Atlas

Click anywhere on the world map and chat with an AI tutor about that place’s history, culture, geography, landmarks, and more. It is an AI-powered geography and history learning companion.

Comment

Inspiration

Ask Atlas helps student learn about the world through place-based exploration: geography, history, culture, landmarks, events, and global data, with AI-guided explanations and supporting sources.

What it does

The student can find the location of an event and pins it on the map, and also can deep dive into the history of the place, or learn more about the place.

It is using LLM to translate user query into a location then use Geoapify to find the coordinate of the location. It also use other data sources such as Wikipedia, Wikidata and World Bank to give the user information about the place.

The frontend is ReactJS with Leaflet Map, backend is Microsoft Agent Framework, there are 3 agents: Location search agent, history timeline agent and Chat Agent. The agents are using the data from Geoapify, Wikipedia, Wikidata and World Bank currently.

How we built it

I used Codex with GPT-5.6 for ideation, planning, implementation, testing, and deployment. I created separate sessions for each area, including Planning, UX, Agent, and Deployment. The process began with brainstorming in the Planning session, followed by UX design, where I built the user experience without a backend. I then moved to the full-stack development session, and finally to the Deployment session, which covered infrastructure options, GitHub workflows, and infrastructure setup.

During the process, I also established regression testing. Whenever a new feature was added, Codex was required to run the regression tests and add test coverage for the new feature.

Challenges we ran into

Codex is really good to translate the raw idea and put the initial scaffolding of the application. The challenges I encountered was when the code has grown, each additional features might break the existing features. To overcome this challenge I asked Codex to create regression testing, and run the regression whenever there is new feature changes.

Accomplishments that we're proud of

I guess I am quite impressed with Codex capability to translate the idea into working code and also iterate the idea into a MVP product.

What we learned

Detail instruction and intention is important to have Codex generate a quality result.

What's next for Ask Atlas

Adding more data sources to have broader information. Add voice feature.

Built With

Share this project:

Updates