Lodestar is an AI-powered burnout prevention companion built for university students in demanding degrees like pharmacy, medicine, nursing, and engineering. It brings together daily journaling, mindfulness support, habit tracking, and career clarity into one caring conversation. This project addresses both the AI-Powered Mindfulness, Journaling and Habit-Tracking Assistant and Agentic Personal Assistant challenge statements.
Lodestar is a conversational AI companion that helps university students prevent burnout before it happens. Students check in daily through a chat interface where Lodestar listens, tracks patterns over time, and responds like a caring friend who actually pays attention.
Most burnout tools focus on symptoms. Lodestar goes one layer deeper. It tracks three things together: wellbeing and burnout risk scored and smoothed across every check-in, a weekly micro-habit the student commits to (which actively feeds back into their burnout score when kept consistently), and career clarity. That last one is the part most apps miss. A major hidden driver of student burnout is not knowing what you actually want and exhausting yourself trying to figure it out by doing everything at once.
When Lodestar notices a student overcommitting out of anxiety rather than direction, it gently names that pattern and helps them hear what they are already saying. It does not prescribe. It reflects.
Under the hood, every message triggers a multi-agent pipeline. A Burnout Radar Agent, a Career Clarity Agent, and a Balance Coach Agent all run in parallel, each looking at the student's message from a different angle. A Daily Companion Agent then pulls their outputs together into one warm, contextual reply. Scores are smoothed over time so a single good day cannot mask a difficult week.
The whole thing runs on Azure, using GPT-4o via Azure OpenAI for conversations, Cosmos DB for storing check-ins and career reflections, and Azure AI Search to surface patterns from past entries. That last part is what lets Lodestar say things like "last time you mentioned feeling disconnected in the lab, is that still true?"
Daily check-ins and journaling Students check in each day and Lodestar responds like a caring friend who actually pays attention. It remembers what was said before and connects the dots over time, surfacing patterns the student might not notice themselves.
Burnout tracking Every check-in is scored and tracked. The wellbeing trend panel shows a weekly chart so students can see how their burnout risk has moved over time, not just how they feel today.
Habit support Students can set one small weekly habit and log it each day. Lodestar celebrates consistency without shaming missed days. The habit panel shows a 7-day dot view and a progress bar.
Career clarity This is the part that makes Lodestar different. A lot of student burnout comes from career anxiety disguised as busyness. Lodestar gently names that pattern when it notices it and helps students hear what they are already saying, without telling them what to do.
- Azure OpenAI (GPT-4o) for all conversations, burnout scoring, and career reflections
- Azure Cosmos DB for storing check-ins, habits, and career profiles
- Azure AI Search for semantic pattern matching across past entries
- Azure App Service for hosting the Node.js backend and serving the frontend
- Backend: Node.js with Express, deployed on Azure App Service
- Database: Azure Cosmos DB for check-ins, habits, and career profiles
- Search: Azure AI Search for semantic pattern matching across past entries
- AI: Azure OpenAI GPT-4o for all conversations and career reflections
- Frontend: HTML, CSS, and JavaScript with no frameworks
From the server, three main flows happen:
1. Chat flow (POST /api/chat)
- Azure AI Search fetches relevant past check-ins as context before anything else runs
- Azure Cosmos DB reads the student's habit and career profile
- Three specialist agents run in parallel: Burnout Radar, Career Clarity, and Balance Coach
- The Daily Companion Agent pulls all three outputs together via Azure OpenAI GPT-4o
- Azure Cosmos DB saves the new check-in
- Azure AI Search indexes the new entry for future pattern matching
2. Career reflection flow (POST /api/career/reflect)
- Azure Cosmos DB reads the existing career profile
- Azure OpenAI GPT-4o runs the career clarity conversation
- Azure Cosmos DB updates the profile with new interests and values
3. Habit flow (GET/POST /api/habit)
- Azure Cosmos DB reads and writes the student's habit and daily check-ins
The Express backend is stateless so it scales horizontally behind Azure App Service without any architectural changes. Cosmos DB partitions by userId so performance stays consistent as the user base grows. Azure AI Search scales independently from the database, keeping search latency stable as more entries are indexed over time.
- All users are identified by anonymised IDs only. No personally identifiable information is collected or stored.
- API keys are managed through environment variables and never exposed in the codebase.
- Azure OpenAI includes built-in content filtering and safety layers across all conversations.
- Visit the live demo at https://lodestar-web.azurewebsites.net/?demo=true
- Live url at https://lodestar-web.azurewebsites.net/
Generative AI was used to assist in developing the code for this project.