Skip to content

StayUp AI

StayUp AI is an AI-driven decision support system that helps coaches manage players and make smarter substitution decisions using real-time risk, fatigue, and performance insights.

Code Video Cerebral Valley

Core trick

Runs a multi-agent risk scorer over player load and injury history, with RAG via Qdrant grounding Gemini answers on substitution risk.

Tags

healthagents

Description

StayUp AI is an AI-driven decision support system that helps coaches manage players and make smarter substitution decisions using real-time risk, fatigue, and performance insights.
Dakshkumar24/STAYUP-★ 00 forksTypeScriptCreated Jan 31, 2026 · last push Feb 1, 2026

README excerpt

# StayUp AI ⚽️🏃‍♂️

**StayUp AI** is an intelligent Decision Support System for sports injury risk management and substitution strategy. It combines **Agentic AI**, **RAG (Retrieval-Augmented Generation)**, and **Predictive Analytics** to help coaches and medical staff make data-driven decisions.

## 🚀 Key Features

*   **Multi-Agent Risk Engine**: A sophisticated backend pipeline (`FeatureBuilder`, `Retriever`, `RiskScorer`, `Readiness`, `Impact`, `Popularity`, `Explainer`) that analyzes player data from multiple angles.
*   **Conversational Assistant**: A built-in chat interface powered by **Google Gemini 2.5** and **Qdrant Vector Search**. Ask natural language questions like *"What is the risk for Fabian Schär?"* or *"How does his injury history compare to others?"*.
*   **Player-Specific Dashboard**:
    *   **Injury Risk Card**: Dynamic risk assessment (Stable, Rising, Critical).
    *   **StayUp Score**: Real-time readiness/availability metric.
    *   **Fatigue Analysis**: Visual fatigue timeline tracking.
    *   **Performance Chart**: Interactive chart to select players and compare stats.
*   **Historical Evidence**: Retrieves similar historical injury cases to justify decision-making (Case-Based Reasoning).

## 🏗 System Architecture

The project is a monorepo divided into two main parts:

### 1. Backend (`/backend`)
*   **Framework**: FastAPI (Python)
*   **AI Models**: Google Gemini 2.5 Flash (`google-generativeai`)
*   **Vector Database**: Qdrant (running in Docke

Team