Inspiration
Currently, 911 text support is limited to a few states, with no real interface to be able to do this in, limiting the kind of help that can be received through this service. We're stepping in to not only provide a GUI that intelligently constructs distress messages based on keywords, but also handle the call to the operator.
What it does
Users in distress can either select from a pre set quick options menu or enter keywords of their own, allowing for flexibility in the content of the distress calls to be generated. The requests will also use context in the form of the User's location and medical history, providing as much background info as possible to aid in the dispatch.
How we built it
The user facing portion was built in React Native, with Tailwind integration for quick styling and Gluestack UI as our component library. The backend was built in ExpressJS, making use of the Gemini, 11Labs, and Twilio APIs.
Challenges we ran into
Gemini at times struggled to understand system prompts, resulting in it constructing responses with unnecessary information unrelated to the type of distress call requested. We got around this with a trial and error approach, experimenting with fine tuning the system prompts until we found one that satisfied our needs.
Accomplishments that we're proud of
We're pleased that we were not only able to get an MVP but also make Veah respond to the operator's request for additional information. We know how crucial this can be for dispatching an appropriate response to the emergency on the operator's side, and are proud to have it be one of the features we can demo to the public.
What we learned
As mentioned before, we are using several APIs to accomplish our goals, and making them all work cohesively proved to be quite a challenge. We are thrilled to report though that our efforts in debugging, trial and error, and just plain persistence payed off to conclude in the product we are able to demo today.
What's next for Veah
To collaborate with local police / fire departments, as well as hospitals to make for better routing of the distress call based on parameters like location, type of call, etc. This will ensure that time between call initiation and help dispatch is cut down, meaning more emergencies handled and more satisfied users.
Built With
- 11labs
- express.js
- gemini
- react
- tailwind
- twilio
Log in or sign up for Devpost to join the conversation.