bloc3_Streamlit / app.py
DavidRambeau's picture
Update app.py
69144ae verified
Raw History Blame Contribute Delete
2.45 kB
# dashboard/test_quality_dashboard.py
import os
import streamlit as st
import pandas as pd
import plotly.express as px
from sqlalchemy import create_engine
from dotenv import load_dotenv, find_dotenv
env_path = find_dotenv()
load_dotenv(env_path, override=True)
# st.set_option('wideMode' , True)
st.set_page_config(
page_title="Dashboard Suivi des indicateurs qualité",
layout="wide"
)
st.header("📊 Dashboard test Video")
st.markdown("""
Ce dashboard présente les indicateurs clés de performance (KPI) des tests automatisés
exécutés sur le modèle de détection de fraude. Il permet de suivre la qualité du modèle
au fil du temps et d'identifier rapidement les éventuels problèmes.
""")
# Connexion DB avec SQLAlchemy
engine = create_engine(os.getenv("BACKEND_STORE_URI"))
# KPIs
col1, col2, col3, col4 = st.columns(4)
# Dernières 24h
df_recent = pd.read_sql("""
SELECT
COUNT(*) as total,
SUM(CASE WHEN status = 'passed' THEN 1 ELSE 0 END) as passed,
AVG(duration_ms) as avg_duration
FROM test_logs
WHERE created_at >= NOW() - INTERVAL '24 hours'
""", engine)
col1.metric("Total Tests (24h)", df_recent['total'].iloc[0])
passed = df_recent['passed'].iloc[0]
total = df_recent['total'].iloc[0]
success_rate = (passed / total * 100) if total > 0 else 0
col2.metric("Taux Réussite", f"{success_rate:.1f}%")
avg_duration = df_recent['avg_duration'].iloc[0] or 0
col3.metric("Durée Moyenne", f"{avg_duration:.0f}ms")
# Graphique évolution accuracy
df_accuracy = pd.read_sql("""
SELECT
DATE(created_at) as date,
AVG(model_accuracy) as accuracy
FROM test_logs
WHERE test_name = 'test_model_accuracy_threshold'
AND created_at >= NOW() - INTERVAL '30 days'
GROUP BY DATE(created_at)
ORDER BY date
""", engine)
fig = px.line(df_accuracy, x='date', y='accuracy',
title="Évolution Accuracy (30 jours)",
labels={'accuracy': 'Accuracy', 'date': 'Date'})
fig.add_hline(y=0.92, line_dash="dash", line_color="red",
annotation_text="Seuil minimum (92%)")
st.plotly_chart(fig)
# Table des derniers échecs
st.subheader("🔴 Derniers Tests Échoués")
df_failures = pd.read_sql("""
SELECT
test_name,
error_message,
created_at,
git_commit_hash
FROM test_logs
WHERE status = 'failed'
ORDER BY created_at DESC
LIMIT 10
""", engine)
st.dataframe(df_failures)