PowerPosition

AI coaching agent for competitive ultimate frisbee players who train without a coach.

What it does

PowerPosition puts a personal coaching agent in your pocket via Telegram. Log training sessions, get pre-game mental prep, and receive post-game check-ins that adapt to your history. After every few sessions the agent updates your personalized coaching plan across short, medium, and long-term goals. A web dashboard shows your training calendar, evolving plan, and a full log of what the agent did and why.

Stack

Layer	Technology
Frontend	React + Vite + TypeScript
Styling	Tailwind CSS v4 + shadcn/ui
Backend	FastAPI (Python)
Agent	LangGraph
Database	MongoDB Atlas
Vector Search	MongoDB Atlas Vector Search
Memory	MongoDBSaver (short-term) + MongoDBStore (long-term)
LLM	OpenAI GPT-4o
Embeddings	OpenAI text-embedding-3-small
Bot	python-telegram-bot
Scheduler	APScheduler

MongoDB usage

• Vector Search — training knowledge base queried on every coaching recommendation
• MongoDBSaver — checkpoints full conversation state per athlete thread
• MongoDBStore — long-term athlete profile (position, goals, key insights)
• Document collections — calendar entries, sessions, coaching plans, agent logs, workflow state

Setup

Prerequisites

• Python 3.11+
• Node 20+
• MongoDB Atlas cluster (M0 free tier is fine)
• OpenAI API key
• Telegram bot token from @BotFather

Environment

Create backend/.env:

MONGODB_URI=mongodb+srv://...
OPENAI_API_KEY=sk-...
TELEGRAM_BOT_TOKEN=...
MONGO_DB_NAME=powerposition
INACTIVITY_THRESHOLD_HOURS=48
SESSIONS_BEFORE_PLAN_UPDATE=3

Install

# Backend
cd backend
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt
playwright install chromium

# Frontend
cd frontend
npm install

Run

# Terminal 1 — backend (FastAPI + Telegram bot)
cd backend && uvicorn main:app --reload --port 8000

# Terminal 2 — frontend
cd frontend && npm run dev

Frontend: http://localhost:5173
Backend: http://localhost:8000

Screens

• Landing — deep space intro with Explore CTA
• Calendar — monthly grid with prev/next navigation, event type tagging (game / training / rest), diary entries, weekly and monthly objectives
• Coaching Plan — three-horizon plan (short / medium / long term), cycle status indicator, agent logs feed with Vector Search sources
• Config — Telegram setup, training material upload, MCP integrations panel

Telegram commands

Command	What it does
/start	Onboarding — sets up athlete profile
/log	Log a training session
/pregame	Pre-game mental prep
/postgame	Post-game check-in (2–3 follow-up questions)

Demo endpoints

curl -X POST http://localhost:8000/api/simulate/pregame
curl -X POST http://localhost:8000/api/simulate/postgame
curl -X POST http://localhost:8000/api/simulate/inactivity