• 

  Home Page

• 

  Dashboard with AI generated Report and suggestion

• 

  Recommened Diet Plan

• 

  Strength Training Plan

About Agada Health

Inspiration

The idea for Agada Health came from my personal struggle to stay motivated in my fitness journey. As someone who has faced weight challenges, I was inspired by how difficult it can be to find a proper diet and strength-training schedule. With Perplexity’s Sonar API, generating tailored plans has become significantly easier.

What it does

Agada Health is an AI-powered coaching platform that seamlessly tracks:

• Profile & Demographics (height, age, sex, goal weight)
• Workouts and Weight History
• Body Metrics (BMI, BFP, BMR, TDEE, etc.)

Through a chat-based interface, it generates:

• Personalized fitness routines
• Strength-training programs
• Diet plans

All tailored to your goals, medical history, and activity level.

How we built it

1. Database & ORM
   • Modeled users, sessions, profiles, weight logs, and AI suggestions in SQLite using Drizzle ORM for type-safe relational data.
2. Backend & API
   • Developed Next.js server functions to handle profile CRUD, weight logging, and calls to Perplexity Sonar API for metric calculations and content generation.
3. AI Prompting
   • Crafted precise system/user prompts and JSON/Markdown schemas to ensure the LLM returns valid, structured outputs every time.
4. Frontend
   • Implemented a chat-driven UI in Next.js where users can log data, ask questions, and receive formatted recommendations in real time.

Challenges we ran into

• LLM Consistency: Ensuring the AI adhered strictly to JSON schemas or Markdown formats required multiple prompt refinements.
• Time & Date Logic: Converting JS Date objects to SQLite timestamps and accurately computing age across time zones.
• Data Integrity: Managing cascading deletes and foreign-key relationships so that deleting a user cleans up all related data.
• Error Handling: Handling occasional API failures or unexpected responses gracefully.

Accomplishments that we're proud of

• End-to-End Integration: A seamless flow from data entry to AI-driven recommendations.
• Structured AI Outputs: Schema-driven prompts that reliably produce valid JSON or Markdown, reducing post-processing.
• Modular Architecture: Clear separation of database models, server logic, and AI layers, simplifying future extensions.
• Personalization at Scale: Automatic tailoring of fitness and nutrition plans based on individual metrics, goals, and restrictions.

Use of Perplexity Sonar API

Body-Metrics Calculation

• calculateBodyMetrics
  Sends user height, age, sex, and weight to Perplexity with a JSON schema prompt
  Returns a JSON object of metrics (BMI, BFP, BMR, TDEE, etc.)

Metric Explanation

• explainBodyMetrics
  Takes a single metric label and full metrics payload
  Requests a Markdown-formatted explanation in six sections:
  
  • measures
    
  • importance
    
  • comparison
    
  • improvement
    
  • tips
    
  • diet

General Coaching Suggestions

• generateSuggestions
  Feeds current vs. goal weight, metrics, medical history, and activity level to Perplexity
  Outputs a JSON list of nutrition, exercise, and lifestyle tips with rationales

Strength-Training Program Generation

• generateStrengthTrainingSuggestions
  Markdown outline

• generateWorkoutFromSuggestion
  JSON schema for full program

Diet-Plan Creation

• generateDietSuggestions
  Sends user context to Perplexity and requests a five-section Markdown diet plan

Input Normalization & Parsing

• generateWeight

• generateProfileInfo
  Extracts or converts free-form text into canonical JSON fields

What we learned

• Proper use of Perplexity Sonar API for structured AI responses
  
• Advanced prompt engineering to deterministic model behavior
  

What’s next for Agada Health

The functionality I have on my mind that I am planning to add after judging period are:

• Sync workout program with google calendar -Ability to track workout inside the application

Built With

• drizzle
• next.js
• perplexity
• plerplexity-ai
• turso