GluCorp AI

• 

  GluCorp AI

Inspiration

My father has diabetes which really got him off-guard. He didn't know how to handle it and finding education and information on diabetes is quite hard considering the web doesn't give specific answers. This led me to create a bot that will answers all questions on diabetes type 1, type 2 and gestational diabetes.

What it does

The bot answers questions related to diabetes in both voice and text. This helps the blind to be able to access and use the chatbot. It also increases efficiency because users can learn as the go due to the audio feature. This is also an educative project which will help educate the masses on matters concerning diabetes making it a social good project.

How we built it

I used the TiDB VectorStore to store the vectors created when the text is embedded. I used Llama to create the storage context which stores the nodes of the text document as well as the vector store index. Finally, I used the Google TTS API for converting text to speech.

Challenges we ran into

Deploying the bot to Streamlit.

Accomplishments that we're proud of

Adding voice feature to the chatbot.

What we learned

How to create a TiDB connection and initializing a TiDB vector store.

What's next for GluCorp AI

GluCorp AI is part of a larger project where we are creating an app that helps detect and manage gestational diabetes. The chatbot will be intergrated in the mobile app to be used by the app users.

Built With

• google-web-speech-api
• llamaindex
• python
• streamlit
• tidb