SportOps: AI-Powered Sports Analysis System

An intelligent multi-agent framework for comprehensive sports analysis using AI, video analysis, statistical research, and medical insights.

---

Table of Contents

• Problem Statement
• Solution Overview
• System Architecture
• Agent Framework
• Setup & Installation
• Project Structure
• Technologies

---

Problem Statement

Modern sports analysis requires synthesizing multiple data sources - video footage, statistical data, research articles, medical records, and performance metrics - into actionable insights. This traditionally requires multiple specialists working across different domains.

Key Challenges:

• Fragmented Data Sources - Videos, statistics, medical records, research articles scattered across platforms
• Time-Consuming Manual Analysis - Domain expertise required across multiple disciplines (biomechanics, medicine, statistics, video analysis)
• Interdisciplinary Perspective - Need for coordinated insights from video analysis, tactical analysis, medical history, and performance metrics
• Scalability Issues - Difficult to scale analysis to handle large datasets and multiple athletes
• Consistency & Quality - Maintaining consistent analysis quality across different specialists and time periods

---

Solution Overview

SportOps is an AI-powered multi-agent system that orchestrates specialized AI agents to collaboratively analyze sports performance from multiple angles. The system leverages:

• Google Gemini AI for intelligent reasoning and analysis
• Multi-Agent Orchestration (Google ADK A2A) for coordinated analysis workflows
• Specialized Sub-Agents for domain-specific expertise (video analysis, data research, medical analysis, etc.)
• Stateful Session Management to maintain context across analysis phases

Key Benefits

✅ Comprehensive Multi-Perspective Analysis - Combines video, statistics, medical, tactical, and biomechanical insights
✅ Intelligent Orchestration - Agents work sequentially and in parallel based on analysis requirements
✅ Scalable Architecture - Modular design enables easy addition of new specialized agents
✅ Context-Aware Processing - Session state tracking maintains continuity across analysis phases
✅ Extensible Integration - Simple API for connecting external data sources and analysis tools

---

System Architecture

High-Level System Architecture

---

Agent Framework

Agent Hierarchy

Root Agent (Sequential Orchestrator)
│
├─ Data Researcher Agent
│  ├─ Stats Researcher Agent (Google Search for statistics)
│  └─ Information Researcher Agent (Web articles, interviews)
│
├─ Performance Analyst Agent
│  ├─ Tactical Analyst Agent (Game strategy analysis)
│  ├─ Performance Analyst Agent (KPI and metrics evaluation)
│  └─ Physio Agent (Physical conditioning analysis)
│
├─ Medical Analyst Agent
│  ├─ Medical Analyst Agent (Medical history & conditions)
│  ├─ Medical Historian Agent (Historical medical background)
│  └─ Biomechanics Agent (Movement & injury analysis)
│
└─ Head Analyst Agent (Synthesis & Reporting)

Agent Responsibilities

Agent	Purpose	Input Sources	Output
Data Researcher	Gather contextual information and statistics	Web search, APIs, databases	Research summaries, statistical context
Performance Analyst	Evaluate athletic performance metrics	Video, statistics, fitness data	Performance insights, KPI analysis
Medical Analyst	Assess medical history and health factors	Medical records, injury history	Medical assessment, health recommendations
Head Analyst	Synthesize all inputs into final report	All previous agent outputs	Comprehensive analysis report
Video Analyst	Analyze video for technique and tactics	Video files	Frame-by-frame analysis, tactical breakdown
Tactical Analyst	Evaluate game strategy and positioning	Video, game stats	Tactical insights, strategy assessment
Biomechanics Agent	Analyze movement patterns	Video, motion data	Biomechanical analysis, injury risk

---

Setup & Installation

Prerequisites

1. Python

• Download Python

2. UV (Astral Framework)

• UV Installation Guide

3. Gemini API

• Gemini API Key Generation Guide

Step 1: Clone the Repository

git clone https://github.com/Jay2219/SportOps.git
cd SportOps

Step 2: Create Virtual Environment

# Create virtual environment
uv -m venv .venv

# Activate on Windows
.\.venv\Scripts\Activate.ps1

# Activate on macOS/Linux
source .venv/bin/activate

Step 3: Install Dependencies

# Install project with all dependencies
uv sync

Step 4: Configure Google AI Access

# Option 1: Using service account credentials
export GOOGLE_APPLICATION_CREDENTIALS="/path/to/service-account.json"

# Option 2: Using API Key
export GEMINI_API_KEY="your-api-key-here"

# Option 3: Using gcloud CLI
gcloud auth application-default login

Step 5: Run the Application

# Run main application
uv run adk web

---

Project Structure

SprtOps/
│
├── main.py                          # Application entry point
├── pyproject.toml                   # Project metadata & dependencies
├── README.md                        # This documentation
├── LICENSE                          # MIT License
│
├── sports_analyst/                  # Main agent framework
│   ├── agent.py                     # Root agent orchestration & runner setup
│   ├── config.py                    # Agent configuration & settings
│   ├── callback_config.py           # Callback handlers for agent lifecycle
│   ├── __init__.py                  # Package initialization
│   │
│   ├── sub_agents/                  # Primary analysis agents
│   │   ├── __init__.py
│   │   ├── data_researcher_agent.py          # Research coordination
│   │   ├── performance_analyst_agent.py      # Performance evaluation
│   │   ├── medical_analyst_agent.py          # Medical assessment
│   │   ├── head_analyst_agent.py             # Synthesis & reporting
│   │   ├── video_analyst_agent.py            # Video analysis
│   │   ├── tactical_analyst_agent.py         # Tactical strategy
│   │   ├── biomechanics_agent.py             # Biomechanical analysis
│   │   ├── stats_researcher_agent.py         # Statistical research
│   │   ├── information_researcher_agent.py   # Information gathering
│   │   ├── medical_historian_agent.py        # Medical history
│   │   ├── physio_agent.py                   # Physiology analysis
│   │   └──
│   └──
│
└── [Additional configuration and cache files]

Key File Descriptions

• agent.py - Defines the root agent as a SequentialAgent that orchestrates all sub-agents with proper callback configuration
• config.py - Contains AgentConfig class with retry policies and conditional execution callbacks
• custom_tools.py - Implements custom tools for the agents (API calls, external integrations)
• callback_config.py - Defines callback handlers for agent lifecycle events
• sub_agents/ - Directory containing all specialized analysis agents

---

Technologies

Core Framework & AI

• Google Gemini API - Large Language Model for reasoning and analysis
• Google ADK (Agent Development Kit) - Multi-agent orchestration framework

Data Processing & Storage

• Python 3.10+ - Primary programming language
• SQLite - Lightweight embedded database (development)