What it does

The bot provides answers and definitions to patients medical queries.

How we built it

The medical bot takes a pre-made prompt and text from the user. It uses that information to send a request to the openAi API. From there we take the AI's response and send that back in chat to the user.

Challenges we ran into

We faced some syntax challenges with some of us coding in our non-native language of python. There were also issues formatting the output because the bot would ramble past its stopping point or not say enough sometimes. Finally the big issue was correctly prompting the ai. From the workshop we knew our disappointing results were stemming from our unclear instructions we were giving the ai.

Accomplishments that we're proud of

One thing we're proud of is how practical our bot is. It was suppose to help with hospital administration and we think it takes a lot of work of their shoulders by having a simple bot that can talk to patients and answer their questions. We are also very proud of the way we could make the bot filter questions. It isn't a perfect filter but it mostly stays on the rails and doesn't stray far from the medical topic.

What we learned

We greatly grew our knowledge of python during this project. Some of us knew little to none going in but are coming out the other side with an entire discord api bot. Along with that we learned many of the openAi applications. Not only can it answer queries and self filter, but it also helps brainstorm product names which we played with until we all settled on the name Eeve.

What's next for Medical bot

A future implementation for medical bot could be voice activated which would have to be taken care of through some additional features. Visual aides could also help with our medical bot to better help patients understand some medical procedures they still struggle to understand. Finally, we think one day it could be more interactive. We think there are a few ways to make the bot do more back and forth conversations by remembering it's previous queries but couldn't get around to it in the time we had.

Built With

Share this project: