test_api.py

"""
test_api.py
===========
Script de prueba para verificar todos los endpoints de la API.

Requiere que la API esté corriendo:
    python app.py

Ejecutar:
    python test_api.py
"""

import json
import random

import numpy as np
import requests

BASE_URL = "http://127.0.0.1:5000"

CLASS_NAMES = [
    "T-shirt/top", "Trouser", "Pullover", "Dress", "Coat",
    "Sandal",      "Shirt",   "Sneaker",  "Bag",   "Ankle boot",
]


def separator(title: str):
    print(f"\n{'─'*60}")
    print(f"  {title}")
    print(f"{'─'*60}")


# ── 1. /health ────────────────────────────────────────────────────────────────
separator("GET /health")
try:
    r = requests.get(f"{BASE_URL}/health", timeout=5)
    print(f"Status HTTP: {r.status_code}")
    print(json.dumps(r.json(), indent=2, ensure_ascii=False))
except Exception as e:
    print(f"❌  Error: {e}")

# ── 2. /model/info ────────────────────────────────────────────────────────────
separator("GET /model/info")
try:
    r = requests.get(f"{BASE_URL}/model/info", timeout=5)
    print(f"Status HTTP: {r.status_code}")
    print(json.dumps(r.json(), indent=2, ensure_ascii=False))
except Exception as e:
    print(f"❌  Error: {e}")

# ── 3. /predict — imagen aleatoria ────────────────────────────────────────────
separator("POST /predict  (imagen aleatoria 784 píxeles)")
try:
    pixels = [random.randint(0, 255) for _ in range(784)]
    payload = {"pixels": pixels}
    r = requests.post(f"{BASE_URL}/predict", json=payload, timeout=10)
    print(f"Status HTTP: {r.status_code}")
    resp = r.json()
    # Mostrar solo top-3 probabilidades para no saturar la consola
    top3 = sorted(resp.get("probabilities", {}).items(), key=lambda x: -x[1])[:3]
    print(f"  Predicción : {resp.get('class_name')}  (id={resp.get('class_id')})")
    print(f"  Confianza  : {resp.get('confidence'):.2%}")
    print(f"  Top-3 probs: {dict(top3)}")
    print(f"  Inferencia : {resp.get('inference_ms')} ms")
except Exception as e:
    print(f"❌  Error: {e}")

# ── 4. /predict — payload inválido (menos de 784) ─────────────────────────────
separator("POST /predict  (validación: 100 píxeles → debe fallar)")
try:
    pixels = [128] * 100
    r = requests.post(f"{BASE_URL}/predict", json={"pixels": pixels}, timeout=5)
    print(f"Status HTTP: {r.status_code}  (esperado 400)")
    print(json.dumps(r.json(), indent=2, ensure_ascii=False))
except Exception as e:
    print(f"❌  Error: {e}")

# ── 5. /predict/batch ─────────────────────────────────────────────────────────
separator("POST /predict/batch  (3 imágenes aleatorias)")
try:
    batch = [[random.randint(0, 255) for _ in range(784)] for _ in range(3)]
    r = requests.post(f"{BASE_URL}/predict/batch", json={"images": batch}, timeout=30)
    print(f"Status HTTP: {r.status_code}")
    resp = r.json()
    print(f"  Total imágenes : {resp.get('count')}")
    print(f"  Inferencia     : {resp.get('inference_ms')} ms")
    for i, result in enumerate(resp.get("results", [])):
        print(f"  [{i}] {result['class_name']:15s}  confianza={result['confidence']:.2%}")
except Exception as e:
    print(f"❌  Error: {e}")

# ── 6. /predict con datos reales de Fashion-MNIST ─────────────────────────────
separator("POST /predict  (imagen real de Fashion-MNIST via sklearn)")
try:
    from sklearn.datasets import fetch_openml
    print("  Cargando 5 muestras de Fashion-MNIST para verificación real...")
    X_real, y_real = fetch_openml(
        "Fashion-MNIST", version=1, return_X_y=True, as_frame=False
    )
    sample_idx = [0, 1000, 5000, 10000, 20000]
    correct = 0
    for idx in sample_idx:
        pixels   = X_real[idx].tolist()
        true_lbl = int(y_real[idx])
        r        = requests.post(f"{BASE_URL}/predict", json={"pixels": pixels}, timeout=10)
        pred_id  = r.json().get("class_id")
        match    = "✅" if pred_id == true_lbl else "❌"
        if pred_id == true_lbl:
            correct += 1
        print(f"  [{idx:>6}] Real: {CLASS_NAMES[true_lbl]:15s} | "
              f"Pred: {CLASS_NAMES[pred_id]:15s} {match}")
    print(f"\n  Correctas: {correct}/{len(sample_idx)}")
except ImportError:
    print("  sklearn no disponible en este entorno para la prueba real.")
except Exception as e:
    print(f"❌  Error: {e}")

print(f"\n{'═'*60}")
print("  Pruebas finalizadas")
print(f"{'═'*60}\n")