Inspiration
The inspiration for this project comes from the headache of being sick and looking for an appointment at the same time. The person who is running back and forth in the bathroom has to open up a bunch of browser windows with every clinic nearby and try to see which one of them has the ability to book appointments online and then go through the process of trying to book appointments online.
What it does
The application chats with a users, reviews symptoms with one AI agent, then simultaneously scans the patient's location for convenient and immediate appointments.
How we built it
We built this by having a dialogue with the patient, understanding their symptomatology (using #Gemini). We give them a little bit of advice, which is handled by one #CrewAI agent, then transition to a second #CrewAI agent that will go on the headless browsers from #BrowserBase to look in their local area for all the URLs associated with clinics. Browserbase scans through all those websites, identifies websites that are offering applicable times, and then surfaces that back to the patient. The entire process is monitored for PII data flows by Weave's guardrails.
Challenges we ran into
The kind of challenges we ran into were several to integrate between Crew and Browserbase, additional challenges from slide. We learned that the application is more complicated by the heterogeneity of the websites' Browserbase interfaces and pulls from.
Accomplishments that we're proud of
Two engineers who did not know one another paired up, modeled a full application flow, made sacrifices and compromises, and sync built the whole thing!
What we learned
We learned how to use Fly, Weave, Browserbase, and Crew!
What's next for PatientHero
What's next for PatientHero is to leverage the insurance information the patient has. That will limit the clinic options. Then we can complete the appointment booking process through the same natural language session.
Log in or sign up for Devpost to join the conversation.