Implement AI-Driven Infrastructure Observability with Prometheus Client and KServe

                      
                     
import os
from prometheus_client import start_http_server, Summary, Gauge
import time
import requests

# Configuration
PROMETHEUS_PORT = int(os.getenv('PROMETHEUS_PORT', 8000))
KSERVE_URL = os.getenv('KSERVE_URL', 'http://localhost:8080/v1/models/my_model:predict')

# Metrics
request_latency = Summary('request_latency_seconds', 'Time spent processing request')
model_prediction_gauge = Gauge('model_predictions', 'Model predictions count')

# Initialize Prometheus metrics server
start_http_server(PROMETHEUS_PORT)

# Core logic for observability
@request_latency.time()
def predict(data):
    try:
        response = requests.post(KSERVE_URL, json=data)
        response.raise_for_status()  # Raise error for bad responses
        model_prediction_gauge.inc()  # Increment prediction count
        return response.json()
    except requests.RequestException as e:
        print(f'Error during request: {e}')
        return None

if __name__ == '__main__':
    while True:
        # Simulated input data for prediction
        input_data = {'instances': [[1.0, 2.0, 3.0]]}
        result = predict(input_data)
        print(f'Prediction result: {result}')
        time.sleep(5)  # Wait before next prediction