MedManager

• 

• 

  here we see the different types of vision blurriness and thanks to what would be a retrieved image, we identified the kind of vision blur.

The inspiration behind MedManager was desiring a personal app I could log my symptoms in and easily display trends to my doctor while also providing me with information each time I log in about the medications I'm taking. I wanted an application that would both inform me of what I'm taking and help me keep track of my medications and side-effects so I could better explain what I was feeling to my doctor.

What it does

MedManager was designed to save medications, medication history, and side-effects. It was meant to serve as both a journal and source of knowledge. It has a profile page where the user would enter their medications, when they started taking them. The Dashboard features updates about other pages and notifications such as when you're in need of a refill based on how long it's been since you started your medication. The next page is for health news. NewsAPI only allows for the broad health category but ultimately instead of using NewsAPI, it would feature a RAG LLM to specially curate news based on the medications used. The next page is Food Interactions with Replicate's untuned LLM. It unfortunately costs money to tune any LLM to specifically respond like a doctor but that would be the ideal scenario. If I were able to pick a chatbot to use, it would have been ChatGPT because of how accurate it is compared to Replicate's LLM. The final page is the Side-Effects Logger which I'm most proud of, it's meant to be the location where user's enter their side-effects. As they enter the data, a RAG agent would retrieve known data about that specific side-effect. In the app in my example it features the different types of blurry vision because describing that to my doctor was difficult without knowing the proper terminology. Ultimately this hopes to help educate the patient with personalized, recent, and relevant data about their medications. A future update for MedManager would be allowing data to be exported from MedManager to the doctor, probably through Epic Charts or some medical charting platform.

How I built it

I utilized Streamlit's documentation and components to put together this webapp, hard coding examples.

Challenges I ran into

I had many plans for this app that ultimately were not implementable in such a short time and ultimately it cost some functionality in the web app and it cost its aesthetic.

What I learned

I learned a great amount about APIs, medical research relating to medicine, working on both front end and backend with Streamlit and utilizing its components. I recognize such a project would have probably been more implementable if I was in a group but for the News API result and the LLM working, I'm satisfied with what I've learned this Shellhacks.

Built With

• llm
• newsapi
• python
• replicate
• streamlit