Emergency
Edgency
On-device emergency-response guidance for civilians and first responders, running fully locally on mobile
Internet connectivity tends to fail exactly when an emergency needs it most. Edgency is a mobile emergency-response assistant that runs entirely on-device: civilians and first responders get step-by-step guidance grounded in WHO and FEMA protocols, with no server call and no API key required once the app is set up.
The right specialist for the emergency
A cardiac, trauma, or general medical emergency routes to MedPsy-1.7B. An earthquake, flood, or storm routes instead to a multimodal Gemma-4 2B that can read photos and video captured at the scene. Either way, a semantic RAG layer pulls the most relevant protocol chunks - 148 of them, indexed across four workspaces - and injects them into the prompt so the response is grounded in an actual guideline, not the model's memory.
Medical -> MedPsy 1.7B -> medical workspace (68 chunks)
Earthquake -> Gemma 4 2B Multimodal -> general workspace (34 chunks)
Flood/Tsunami-> Gemma 4 2B Multimodal -> water workspace (29 chunks)
Storm -> Gemma 4 2B Multimodal -> storm workspace (17 chunks)One profile, every system prompt
A user's role, experience level, and declared medical conditions are stored on-device and woven into every system prompt, so the same core adapts its guidance to a civilian or a veteran first responder. Voice messages are transcribed on-device with Whisper before reaching the model, and a built-in location tool lets the model give location-aware instructions without any of it leaving the phone.
Demo video
Why the right model gets the right emergency
MORE FROM THE SHOWCASE
Build this
Ship your own local AI
One install in JavaScript or Python. 10+ AI tasks. Private, offline and free - on the hardware you already own.