# Forewarn

> Forewarn is a free early warning app for Africa, floods first. Put a pin on your street, farm or village, and an AI agent tells you if flooding looks unlikely, possible or likely in the next 14 days, with the chance in plain words ("about 1 in 4 forecasts"). Code computes every number from 51 ECMWF forecast runs; the agent explains them, and a checker holds every number it says to the data. Built on Meta's Muse Spark and SAM 3.1 for the Meta Global AI Developer Hackathon (Indie track), October to November 2026.

Floods are where it starts; the same system can watch for drought and storms next. Forewarn works anywhere in Africa, with no account. Lagos and Nairobi also get a replay of the last big flood over the satellite map, street by street. Alerts by email are drafted by an agent that checks each followed area every 6 hours, and a person approves each alert before it is sent. We replayed 36 past African floods with today's rules: 29 were flagged in advance (81%), 28 of them 3 days or more ahead, and Nairobi's flood of 23 April 2024 was flagged 7 days ahead.

## Pages

- [Forewarn](https://forewarn.site/): find your place and get the answer for the next 14 days
- [The film](https://forewarn.site/forewarn-film.mp4): 3 minutes, the Mathare flood of 23 April 2024 and Forewarn at work
- [The time machine](https://forewarn.site/?past): what Forewarn would have said before 36 real floods, 7 days to 1 day ahead, with the scorecard

## Full text

- [llms-full.txt](https://forewarn.site/llms-full.txt): what Forewarn does, how the agent works, the models, the data, the test results, privacy and limits

## Data you can read

- [Scorecard (JSON)](https://forewarn.site/api/past): every past flood, the forecast made 1 to 7 days before, the level of each day and whether an alert would have gone out
- [One flood (JSON)](https://forewarn.site/api/past/0): the same for one flood; numbers run from 0 to 39

## Code

- [github.com/euphoriawolf/forewarn](https://github.com/euphoriawolf/forewarn): the full source, the README, and the test scripts in evals/

## Optional

- [Flood Check](https://flood-check.dawnstackai.com): the street-level flood model for Lagos and Nairobi that Forewarn reads (built before the hackathon)
- [Derrick Ikenga](https://derrickikenga.com): the builder