ChronicCareIQ

Inspiration

Many people with chronic conditions struggle to find reliable, personalized guidance. Generic wellness apps don’t account for individual health profiles, emerging research, or lifestyle constraints. We wanted to build a coach that combines cutting-edge clinical studies, vetted nutrition data, adaptive planning, and real-time customization—so users get truly trustworthy, actionable advice.

What it does

ChronicCareIQ generates a 14-day wellness plan tailored to any chronic condition and individual preferences. Each day combines:

Meals built from up-to-date USDA nutrient data and clinical trial insights Primary activity (e.g. yoga, strength training) Complementary practice (meditation, breathwork) with rationale

After the plan is generated, users can:

Edit any item via chat (“Skip salmon dinner,” “Suggest breathwork instead of meditation,” “Replace breakfast on day 3 with oatmeal”). Mark busy days, and the system will automatically adjust or skip workouts/ give easy prep meals for those dates.

All edits apply instantly, and the plan re-renders to reflect the user’s needs in real time.

How we built it

Frontend: HTML, CSS, JSfor a survey form, two-week grid UI, and chat pane. Backend: FastAPI wrapping LangChain tools as “agents” (Survey, PlanGeneration, Nutrition, Recipe, Wellness, Research, ChatAdjustment). RAG Sources:

USDA FoodData Central API for nutrition facts Spoonacular API for recipes PubMed/ClinicalTrials.gov E-utilities for recent trial abstracts LLM: Microsoft Phi 4 (via LangChain) orchestrates parallel tool calls, synthesizes the initial plan, and handles dynamic plan edits.

Challenges we ran into

API Rate Limits: Caching was needed to avoid throttling during demos. Data Consistency: Merging JSON from multiple sources required a consensus layer to filter weak signals. Real-time Plan Edits: Crafting prompts that reliably produce structured JSON diffs took several iterations. Busy-day Adjustments: Designing logic so the plan gracefully skips or reshuffles entries on user-marked busy days.

Accomplishments we’re proud of

Full 14-day Plan MVP delivered in under 48 hrs, complete with diet + wellness pairings and real-time editability. Live RAG Integration from three distinct APIs, demonstrating truly data-driven recommendations.

• Chat-First Editing letting non-technical users personalize their plan naturally, including marking busy days for automatic adjustments.

What we learned

• Credibility Matters: Inline source citations and consensus checks dramatically boost trust.
• Modular Agents: LangChain’s tools-and-agent pattern allowed rapid stubbing and iteration.
• Prompt Refinement: Clear instructions are key to reliable JSON patch outputs for edits.
• User Flexibility: Giving users control to tweak meals and workouts (or skip days) is essential for real-world usefulness.

What’s next for ChronicCareIQ

• Expand Sources: Integrate wearable-sensor data (heart rate, sleep) and local grocery availability for shopping lists.
• Community Features: Gamified challenges, peer Q&A, and habit sharing to boost adherence.
• Mobile PWA: Offline-capable lightweight app for on-the-go access.
• Clinical Validation: Pilot with real users and add telehealth provider feedback loops.
• Predictive Adjustments: Auto-detect upcoming calendar conflicts and proactively adjust the plan.

Built With

• css
• fastapi
• github
• html
• javascript
• langchain
• langgraph
• python