AfiaPlus

Inspiration

Language barrier and lack of comprehensive medical and healthcare information and access to reliable diagnostic tools are some but of many challenges community health care workers face in Kenya. Many community health workers lack tools and information to make informative decisions while on their day to day activities.

AfiaPlus aims to solve this problems using a Chat Bot and Symptoms Checker using Vector Search for RAG leveraging data from local medical health practitioners and researchers. With collaboration between health practitioners,researchers and tech, leveraging LLMs and vector databases AfiaPlus aims to be the ultimate tool for community health care workers in bridging the gap in access to medical and health information.

What it does

AfiaPlus health chat bot and symptoms checker provide community health workers with quick access to updated medical information that can be translated to local languages. Symptoms checker provides community health workers with a reliable tool to help them perform quick diagnosis that helps give actionable information.

Leveraging vector databases, medical health professionals and researchers provides data for up to date health and community related knowledge bases for to provide context for Gemini

Data is gathered from medical health practitioners and researchers converted to vectors using Gemini API for retrieval and similarity search to be used as context with Gemini API .

When a user interacts with either Chat Bot or Symptoms Checker, data is retrieved by performing similarity searches in the vector database and used as context with Gemini API .

How we built it

• Front-end Built with Angular
• Back-end Built using Python and Flask api.
• Database Used TiDB cloud serverless for data management, vector search and retrieval.
• Hosting Vercel and Render for hosting and deployment.
• LLMs Used Gemini API for chat and content generation.
• Vector Embeddings Gemini API for vector embeddings generation.

Challenges we ran into

• Lack of consistency in Gemini API responses.
• Lack of local up to date medical and health related journals.
• Gemini API at times hallucinates when translating between local dialects, even at times request context is lost during translation.
• Using RAG with the vector database required management of data retrieval and embedding processes, and optimizing performance to ensure timely and accurate itinerary generation was quite challenging.
• Uploading and managing large datasets was a challenge.
• Timely and relevant responses was challenging due to lack of relevant documents.

Accomplishments that we're proud of

This being my first chat bot, am very proud of seamless integration between Gemini API , vector databases and how they can both be used together.

What we learned

Thought the project i have managed to solidify knowledge on use of Gemini API , vector databases, etc and gained insights and challenges involved in developing and deploying AI driven apps.

What's next for AfiaPlus

I will continue to update AfiaPlus to add more features.

Built With

• angular.js
• flask
• onrender
• python
• tidb
• vercel