MediAssist

Inspiration

Millions of people lack access to reliable health information, especially outside business hours or in underserved areas. People often turn to unreliable sources when concerned about symptoms. MediAssist was built to provide trustworthy, conversational health guidance — always encouraging professional consultation for serious concerns.

What it does

MediAssist is a conversational AI agent that helps users understand symptoms, get medication information, and receive personalized health guidance. Most importantly, it remembers you across sessions — if you mentioned your allergies last week, it knows this week too.

How we built it

Qwen AI (qwen-plus) via Qwen Cloud for natural language health guidance
MongoDB Atlas for persistent user profiles, symptom logs, and chat history
Streamlit for the web chat interface with sidebar history navigation
Alibaba Cloud ECS (Singapore) for cloud deployment
Python 3.9 as the backend language

Challenges we ran into

Deploying on CentOS 7 required compiling Python 3.9 from source and resolving complex dependency conflicts with pyarrow and pandas
Designing the dual memory context system to maintain agent memory even when users start new conversations
Resolving Streamlit recursion depth issues specific to our server configuration

Accomplishments that we're proud of

Built a fully functional deployed AI health agent with true persistent memory
Created a polished ChatGPT-style interface with sidebar chat history
Successfully deployed on Alibaba Cloud ECS with free trial credits
Agent remembers users across sessions, devices, and new conversations

What we learned

Qwen Cloud API integration using OpenAI-compatible SDK
MongoDB Atlas collection design for efficient chat history retrieval
Alibaba Cloud ECS deployment, security groups, and persistent process management with systemd
Streamlit session state management for complex multi-session AI applications