This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass I built **HealthTuner ** for my friend Alex, an ambitious fitness enthusiast who constantly balances heavy training sessions with a demanding desk job. Alex was struggling with burnout and plateauing because tracking workouts across fragmented note apps failed to give any real, holistic insight into his volume, recovery markers, sleep quality, and nutrition.
HealthTuner solves this by providing a streamlined, full-stack health tracking and AI companion platform:
Interactive Calendar Dashboard: An intuitive, month-at-a-glance calendar with smart date access controls (active past/present days, protected future dates) and workload color badges.
Dynamic Multi-Tier Logging: Adapts dynamically to how much time the user has:
Easy: 1 rapid open-input check-in.
Medium: 3 structured inputs (Activity, Nutrition/Hydration, Energy Notes).
Maximum: 5 deep inputs (Training Breakdown, Macros/Diet, Sleep/Recovery, Mental Stress & Wellness, Key Wins).
AI Health Intelligence & Deload Advisory: Ingests multi-day logs and generates a structured, two-part analysis featuring behavioral summaries and prioritized, actionable deload/recovery recommendations.
Dark/Light Mode & Mobile-Ready Experience: Tailored for both desktop and on-the-go gym logging.
Live Application: https://health-tuner.vercel.app Live Database: Powered by Neon Serverless PostgreSQL with pooled connections.
GitHub Repository:
HealthTuner is an intelligent, full-stack Health & Workout Companion Web Application built with FastAPI, Neon Cloud PostgreSQL, SQLAlchemy, Jinja2, and Tailwind CSS, integrated with Backboard.io for structured AI health intelligence and recommendations.
Easy, Medium, Maximum).
HealthTuner is architected with a modern, reliable, and scalable open-source stack:
Open-Source AI & Backboard.io SDK: Integrated Meta-Llama-3-8B-Instruct via Backboard.io for fast, structured behavioral analysis and health coaching.
Built a custom pipeline that synthesizes messy workout notes, macro targets, and sleep metrics into structured markdown summaries with clear headings and bulleted action plans.
Included an intelligent heuristics fallback engine to ensure high availability and zero downtime.
Backend & Cloud Database:
FastAPI (Python 3.12): High-performance ASGI framework powering auth, REST APIs, and dynamic Jinja2 server-side rendering.
Neon Serverless PostgreSQL: Cloud database with instant branching and pooled connections for rock-solid data persistence across workouts.
SQLAlchemy ORM & Bcrypt: Secure data modeling, cascade relationships, and password encryption.
Frontend & Polish:
Tailwind CSS & Vanilla CSS: Clean, responsive layout featuring custom micro-animations and zero-flash dark/light mode switching.
Marked.js & Lucide Icons: Rich typography rendering for AI recommendations and crisp iconography.
Open innovation and open-weight models (like Llama 3) fundamentally democratize access to personalized health coaching. In traditional closed ecosystems, sensitive health, lifestyle, and biometric logs are locked behind proprietary, expensive paywalls and opaque APIs.
By leveraging open models and open-source infrastructure:
Full Transparency & Privacy: Users maintain control over where their data goes and how health summaries are generated.
Agility & Customization: Prompts, fine-tuning, and inference pipelines can be adjusted locally or migrated across providers without vendor lock-in.
Accessibility: Powerful AI coaching is made freely available to friends, athletes, and anyone wanting to lead a healthier life without subscriptions.
Best Overall Project
Neon Cloud Postgres / Database Category
Backboard.io AI & Open Models Category