AllanOnyonka-ltsm/WildFire — reverse-engineered prompt

Build me a simple AI wildfire detection web app in Python.

I want it to open in the browser with a clean Streamlit interface where someone can either upload an image or choose a location using latitude and longitude, then check recent NASA satellite imagery for possible wildfire signs. The app should use a MobileNet based model saved in a models folder, show a clear fire probability result, and let the user adjust the confidence threshold.

For the satellite part, fetch NASA GIBS imagery for the selected point over the past few days, analyze each day, and show the images with their results. Include useful details like inference time and prediction confidence, but keep the interface understandable for a normal user.

Please make it easy to run with requirements.txt and streamlit run main.py. If the model file is missing, show a friendly error explaining where it should go. Look up current docs online if needed.