AI Workout Coach

AI Workout Coach is your friendly, chat-based fitness planner that creates personalized workout routines in seconds. Just tell your goal, gear, and schedule and your AI coach takes care of the rest.

Comment

Inspiration

Most fitness apps are either too rigid or overwhelming. People want expert workout guidance without needing to hire a personal trainer or sift through endless articles and YouTube videos. We wanted to simplify that experience using a conversational AI that feels like chatting with a coach — fast, friendly, and personalized.

What it does

AI Workout Coach is a chat-based fitness planner that:

  • Greets users like a friendly personal trainer
  • Asks for their fitness goals, available equipment, days per week, and session duration
  • Generates a fully structured, multi-day workout plan using DeepSeek R1
  • Responds only to fitness-related questions (and politely declines anything else)
  • Stores past plans in a side panel so users can revisit them anytime

All through a clean, mobile-friendly interface built with shadcn/ui.

How we built it

  • Frontend: React (with shadcn/ui for styling), chat interface styled like ChatGPT
  • Backend API: DeepSeek R1 via Fireworks AI for intelligent workout generation
  • Routing: React Router for navigation between landing, auth, and dashboard
  • State: React state + localStorage for persistence
  • Bolt Platform: Built in a single-shot using Bolt.new with declarative and code blocks

Challenges we ran into

  • Designing a natural conversation flow for a workout coach using only a single API call per input
  • Validating user input in a chat-style interface while keeping the UX smooth
  • Limiting the AI's scope strictly to fitness questions to avoid irrelevant responses
  • Fitting all logic, UI, and backend code into a one-shot prompt format without follow-ups

Accomplishments that we're proud of

  • Created a fully functional, friendly AI fitness coach with personalized workouts in real-time
  • Seamlessly integrated chat UX and structured plan formatting
  • Maintained scope and control over AI outputs (fitness-only)
  • Delivered everything in one-shot under hackathon constraints

What we learned

  • How to engineer a tightly scoped prompt that maximizes value from an LLM while preventing unwanted drift
  • DeepSeek R1 performs excellently for structured task-based completions like fitness plans
  • Chat-based interfaces improve user engagement, even for utility-driven apps
  • Bolt's one-shot architecture forces product clarity and rapid iteration

What's next for AI Workout Coach

  • Add image previews for exercises
  • Enable calendar integration for workout reminders
  • Support progress tracking with visual charts
  • Offer nutrition tips based on fitness goals
  • Expand login with OAuth and cloud database (e.g., Supabase)
  • Launch to early users via Product Hunt, Reddit, and fitness communities

Built With

Share this project:

Updates