Tribble

Inspiration

Conflict zones produce a flood of fragmented information: news, satellite imagery, weather, and reports from people on the ground. NGOs rely on outdated briefings; individuals rely on scattered WhatsApp groups and social media. We wanted a single, live view of what’s happening, for escape plans and for aid delivery.

What it does

Tribble, by using a spatiotemporal ontology, turns fragmented crisis data into one live map. Tribble used LCAUD as our core conflict dataset, combined with satellite and weather data to show how regions change. People in conflict zones submit reports; our agents verify them against LCAUD and other sources and update the map in real time. Individuals use it to plan escapes and find resources; NGOs use it to organise aid.

How we built it

We built two pipelines: an automated intelligence layer powered by LCAUD, plus satellite and weather data (and future drone data), and an individual submissions layer. We focused on Sudan for this hackathon. Agents analyse, verify, and merge both streams into a shared ontology. The frontend uses Next.js and Mapbox; the backend uses FastAPI and LangGraph. Supabase and PostGIS handle storage and geospatial queries.

Challenges we ran into

Integrating LCAUD with satellite and weather sources meant dealing with different formats and update frequencies. Balancing speed with verification for individual reports took several iterations. We also had to design an ontology flexible enough for both NGOs and individuals without overwhelming either.

Accomplishments that we're proud of

We shipped a working demo for Sudan that combines LCAUD, satellite, weather, and human-sourced data into a single map. We built a verification pipeline that cross-references reports instead of discarding them. We designed for both individuals and NGOs from the start. We proved the concept in one conflict zone, and we’re motivated to scale it to every conflict.

What we learned

Crisis intelligence is about trust and speed. People need information fast, but verification matters. We learned that “downgrade, don’t discard” keeps more useful signal than strict filtering. We also learned how important it is to design for people in conflict zones, not only for NGO operators. Sudan showed us the need is real and the architecture works.

What's next for Tribble

We’ll expand beyond Sudan to other conflict zones, our goal is to build this for every conflict. We’ll add drone data to the automated pipeline and WhatsApp intake for individual submissions. We’ll add real-time push updates. We’re exploring partnerships with NGOs for pilot deployments. One map, one ontology, for every crisis.