The AI-Powered Coaching Journal for Elite Athletes and Creators.
RepLog is a high-performance web application designed for the IBM WatsonX Hackathon 2026. It transforms raw practice logs into actionable coaching intelligence, powered by IBM's industry-leading Granite Foundation Models.
In a world of data overflow, athletes and creators often struggle to turn "practice volume" into "performance progress." RepLog bridges this gap by acting as a 24/7 AI coach. By analyzing training logs, sensory feedback, and performance metrics, RepLog provides structured, psychological, and technical feedback to help users achieve their "Peak Rep."
RepLog features a bespoke Glassmorphic UI designed for focus and flow:
- Ultra-Modern Aesthetics: Translucent layers with dynamic blur effects and vibrant gradients.
- Micro-Animations: Subtle transitions that provide tactile feedback without distraction.
- Mobile First: Responsive architecture that works on the track, in the gym, or at the desk.
- Granite-Powered Insights: Leveraging
ibm/granite-3-8b-instructto perform domain-specific analysis across sports, coding, and music. - Structured Coaching Loop: Every response follows a high-impact pattern:
- 🎉 Small Wins: Immediate positive reinforcement.
- 🔍 Pattern Observation: Identifying technical or mental trends.
- 🎯 Next Practice Focus: Concrete, actionable steps for the next session.
- Semantic Auto-Titling: Bob analyzes the start of your session and automatically generates a title to keep your history organized.
- Parallel Threads: Maintain separate coaching sessions for "Cardio," "Strength," and "Skill Work" simultaneously.
- Persistence: For registered users, sessions are preserved across reloads with optimized local caching.
- Session Management: Full control to resume, jump back to history, or delete old sessions.
- Drag-&-Drop Intelligence: Seamlessly drag coaching advice or previous logs directly into your persistent journal.
- Read-Only for Guests: Default welcome guides are provided as read-only templates to ensure system integrity.
- User-Scoped Storage: Implements
prefix_userEmailstorage logic to ensure multi-user data isolation on shared machines.
- React (Vite): Lightning-fast HMR and build times.
- Vanilla CSS: Custom design system built from the ground up (no generic frameworks).
- Lucide Icons: Crisp, medical-grade iconography for professional appearance.
- React-Markdown: Full support for GFM-enhanced coaching summaries.
- Node.js / Express: Secure proxy layer to mask API credentials.
- IBM WatsonX AI SDK: Direct integration with
@ibm-cloud/watsonx-ai. - Robust Parsing Engine: Custom regex-based JSON extractor that handles truncated or malformed LLM responses gracefully.
- Node.js v18 or higher.
- IBM Cloud Account with a WatsonX.ai project instance.
Create a .env file in the root directory:
IBM_CLOUD_API_KEY=your_ibm_api_key
IBM_PROJECT_ID=your_watsonx_project_id
PORT=3001# 1. Install dependencies
npm install
# 2. Launch the AI Backend (Terminal 1)
node server/index.js
# 3. Launch the Frontend (Terminal 2)
npm run dev- 📂 App Architecture: Central state and session management.
- 📂 AI Pipeline: IBM WatsonX integration and prompt engineering.
- 📂 Notebook System: Persistent journaling and drag-drop logic.
- Innovation: Real-time conversion of unstructured practice logs into structured coaching plans.
- Reliability: Implements robust error handling for AI inference and secure authentication workflows.
- Scalability: Designed for multi-user expansion while maintaining strict data privacy via user-scoped storage.
Created with 🦾 by Team Vireon for the IBM WatsonX Hackathon 2026. "Optimize every rep. Log every gain."