graph LR
    A[Live Video Stream] -->|OpenCV+YOLOv8| B[Object Detection]
    C[Telemetry API] --> D[Event Normalizer]
    B --> E[Frame‑level Features]
    D --> E
    E --> F[Feature Store (Redis)]
    F --> G[LLM Prompt Builder]
    G --> H[LLM Inference (4‑bit quantised)]
    H --> I[Insight Formatter]
    I --> J{Publish}
    J -->|Slack| K[Bot]
    J -->|Web UI| L[Dashboard]

curl -H "X-API-Key: YOUR_KEY" \
     "https://api.optasports.com/v1/events?match_id=12345&delay=300"

import cv2
import torch
from pathlib import Path

# Load YOLOv8 (small) model, quantised to 4‑bit for speed
model = torch.hub.load('ultralytics/yolov8', 'yolov8s', pretrained=True).to('cuda')
model.half()                     # FP16
model.quantize(bits=4)           # 4‑bit inference

cap = cv2.VideoCapture('rtmp://live.fifa.org/2026_match')
while cap.isOpened():
    ret, frame = cap.read()
    if not ret:
        break
    # Resize to 720p for a good trade‑off
    frame = cv2.resize(frame, (1280, 720))
    results = model(frame)       # returns boxes, confidences, class ids
    # Extract ball (class 0) and players (class 1‑10)
    detections = results.xyxy[0].cpu().numpy()
    # Push detections to Redis for the next stage
    redis_client.publish('detections', json.dumps(detections))

def build_prompt(detections, recent_events):
    ball = detections['ball']
    attackers = detections['players']['attacking']
    defenders = detections['players']['defending']
    last_event = recent_events[-1]

    prompt = f"""You are a football analyst.  
Current ball position: x={ball['x']:.1f}, y={ball['y']:.1f}.  
Attacking players near the ball: {', '.join(attackers)}.  
Defending players nearby: {', '.join(defenders)}.  
Last event: {last_event['type']} at minute {last_event['minute']}.  

Give a concise tactical insight (max 2 sentences) and a probability (0‑100%) that the next attack will result in a goal."""
    return prompt

import transformers

tokenizer = transformers.AutoTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.1")
model = transformers.AutoModelForCausalLM.from_pretrained(
    "mistralai/Mistral-7B-Instruct-v0.1",
    device_map="auto",
    torch_dtype=torch.float16,
    load_in_4bit=True,
)

def get_insight(prompt):
    inputs = tokenizer(prompt, return_tensors="pt").to('cuda')
    output = model.generate(**inputs, max_new_tokens=60, temperature=0.7)
    return tokenizer.decode(output[0], skip_special_tokens=True)

import requests, json

def post_to_slack(text):
    webhook = "https://hooks.slack.com/services/XXX/YYY/ZZZ"
    payload = {"text": text}
    requests.post(webhook, data=json.dumps(payload))

def post_to_telegram(text):
    token = "123456:ABC-DEF1234ghIkl-zyx57W2v1u123ew11"
    chat_id = "-1001122334455"
    url = f"https://api.telegram.org/bot{token}/sendMessage"
    requests.post(url, data={"chat_id": chat_id, "text": text})

def on_insight(msg):
    insight = msg['data']
    post_to_slack(insight)
    post_to_telegram(insight)

redis_client.subscribe(**{'insights': on_insight})