Paani Kahan
- stack
- Hackathon · Geospatial · Agents · AWS
- links
- Code ↗
What it is#
Paani Kahan forecasts which streets will flood in the next six hours and how deep, for every street around you. It checks your saved commute at the time you would reach each street, finds a dry route and opens it in Google Maps. It was built for Environmental Hacks 2026 (Track 02: Heat and Water).
Why it exists#
Flood warnings are usually about a whole city. What matters when leaving the house is whether your own street and your own route are passable. The app works like a maps app, on a phone or a laptop, and stays silent until a new forecast puts your route under water.
How it works#
The map is cut into tiles. Each tile fetches OpenStreetMap streets, terrain from the AWS Registry of Open Data and rain from Open-Meteo, then runs a flood model: rain the drains cannot take runs downhill and fills the low spots. What is on your screen loads first, then your route, then streets near you, then main roads out to 100 km. Drain capacity is tuned to Chennai from 35 years of rain data, and residents' flood reports correct the forecast nearby for 90 minutes.
The AWS open-source tools each have a job: OpenSearch for search, Strands Agents SDK for the agent, Cedar for who may see what, AWS SAM CLI and LocalStack for the alert pipeline, and Finch for containers.
Stack#
Python, OpenSearch, Strands Agents SDK, Cedar, AWS SAM CLI, LocalStack, Finch, OpenStreetMap, Open-Meteo.
Status#
A hackathon build. The README is open about limits: it is calibrated to Chennai's rainfall and checked against past storms, not validated street by street; elevation data is coarse in dense areas; and the demo storm is scripted so it can be shown on a dry day.