Dependencies fix

README.md

@@ -11,7 +11,15 @@ license: mit
 
 # Under Construction
 
+py -3.9 -m venv .venv
 
+.venv\Scripts\activate
+
+python -m pip install --upgrade pip
+
+pip install -r requirements.txt
+
+uvicorn app.main:app --reload
 
 # Repo Structure
 

app/api/routes.py

@@ -6,6 +6,7 @@ from app.inference.predictor import Predictor
 from app.core.logging import log_prediction
 from app.monitoring.data_loader import load_production_data
 from app.monitoring.drift import run_drift_check
+import pandas as pd
 
 
 router = APIRouter()
@@ -38,3 +39,12 @@ def run_drift():
         "status": "drift_check_completed",
         "report_path": report_path
     }
+
+@router.get("/monitoring/run")
+def monitoring_run():
+    # Example: load some data
+    current_data = pd.read_csv("data/current.csv")
+    reference_data = pd.read_csv("data/reference.csv")
+
+    alerts = run_drift_check(current_data, reference_data, model_version="v1")
+    return {"alerts": alerts}

app/monitoring/alerts.py

@@ -1 +0,0 @@
-# threshold evaluation

app/monitoring/drift.py

@@ -4,6 +4,7 @@ import os
 import pandas as pd
 from evidently.report import Report
 from evidently.metric_preset import DataDriftPreset
+from app.monitoring.governance import Governance
 
 REFERENCE_DATA_PATH = "models/v1/reference_data.csv"
 REPORT_DIR = "reports/evidently"
@@ -27,3 +28,20 @@ def run_drift_check(current_df: pd.DataFrame):
     report.save_html(REPORT_PATH)
 
     return REPORT_PATH
+
+# Thresholds configuration
+thresholds = {
+    "psi": 0.2,
+    "accuracy_drop": 0.05,
+    "f1": 0.7
+}
+
+governance = Governance(thresholds=thresholds)
+
+def run_drift_check(current_data, reference_data, model_version="v1"):
+    report = Report(metrics=[DataDriftPreset()])
+    report.run(current_data=current_data, reference_data=reference_data)
+
+    # Governance check
+    alerts = governance.check_metrics(report.as_dict(), model_version=model_version)
+    return alerts

app/monitoring/governance.py

@@ -0,0 +1,66 @@
+# This file implements threshold checking, governance signals logging, and notifications.
+
+import json
+import logging
+from datetime import datetime
+from app.utils.alerts import send_email_alert, send_slack_alert
+import os
+
+os.makedirs("logs", exist_ok=True)
+logger = logging.getLogger("governance")
+logger.setLevel(logging.INFO)
+handler = logging.FileHandler("logs/governance_alerts.log")
+formatter = logging.Formatter('%(asctime)s | %(levelname)s | %(message)s')
+handler.setFormatter(formatter)
+logger.addHandler(handler)
+
+
+class Governance:
+    def __init__(self, thresholds: dict):
+        """
+        thresholds example:
+        {
+            "psi": 0.2,
+            "accuracy_drop": 0.05,
+            "f1": 0.7
+        }
+        """
+        self.thresholds = thresholds
+
+    def check_metrics(self, report_dict: dict, model_version: str):
+        alerts = []
+
+        # Example: data drift
+        psi = report_dict.get("metrics", {}).get("DataDriftPreset", {}).get("result", {}).get("dataset_drift", 0)
+        if psi > self.thresholds.get("psi", 0.2):
+            alerts.append(f"Data drift detected (PSI={psi})")
+
+        # Example: classification performance
+        f1 = report_dict.get("metrics", {}).get("ClassificationPreset", {}).get("result", {}).get("f1_score", 1.0)
+        if f1 < self.thresholds.get("f1", 0.7):
+            alerts.append(f"F1 drop detected (F1={f1})")
+
+        # Example: regression accuracy
+        accuracy_drop = report_dict.get("metrics", {}).get("RegressionPreset", {}).get("result", {}).get("accuracy_drop", 0)
+        if accuracy_drop > self.thresholds.get("accuracy_drop", 0.05):
+            alerts.append(f"Accuracy drop detected ({accuracy_drop})")
+
+        # Log alerts
+        for alert in alerts:
+            self.log_alert(alert, model_version)
+
+        # Optional notifications
+        for alert in alerts:
+            send_email_alert(alert)
+            send_slack_alert(alert)
+
+        return alerts
+
+    @staticmethod
+    def log_alert(message: str, model_version: str):
+        log_entry = {
+            "timestamp": datetime.utcnow().isoformat(),
+            "model_version": model_version,
+            "alert": message
+        }
+        logger.info(json.dumps(log_entry))

app/utils/alerts.py

@@ -0,0 +1,28 @@
+# Helper functions for sending notifications.
+
+import smtplib
+from email.message import EmailMessage
+import requests
+
+def send_email_alert(message: str):
+    # Configure your SMTP settings here
+    try:
+        email = EmailMessage()
+        email.set_content(message)
+        email["Subject"] = "ML Governance Alert"
+        email["From"] = "ml.alerts@example.com"
+        email["To"] = "ops-team@example.com"
+
+        with smtplib.SMTP("localhost") as smtp:
+            smtp.send_message(email)
+    except Exception as e:
+        print(f"Failed to send email alert: {e}")
+
+
+def send_slack_alert(message: str):
+    # Slack webhook URL
+    webhook_url = "https://hooks.slack.com/services/XXXX/YYYY/ZZZZ"
+    try:
+        requests.post(webhook_url, json={"text": message})
+    except Exception as e:
+        print(f"Failed to send Slack alert: {e}")

requirements-dev.txt

@@ -1,5 +1,12 @@
 evidently==0.4.15
-fastapi
-uvicorn
-pandas
-scikit-learn
+fastapi>=0.100.0,<0.130.0
+uvicorn>=0.21.1,<0.40.0
+pandas>=1.5.0,<2.0.0
+scikit-learn==1.6.1
+pydantic==1.10.12
+plotly
+numpy<2.0.0
+requests
+scipy>=1.10.0,<2.0.0
+python-multipart>=0.0.6
+typing-extensions>=4.0.0

requirements.txt

@@ -1,5 +1,12 @@
 evidently==0.4.15
-fastapi
-uvicorn
-pandas
-scikit-learn
+fastapi>=0.100.0,<0.130.0
+uvicorn>=0.21.1,<0.40.0
+pandas>=1.5.0,<2.0.0
+scikit-learn==1.6.1
+pydantic==1.10.12
+plotly
+numpy<2.0.0
+requests
+scipy>=1.10.0,<2.0.0
+python-multipart>=0.0.6
+typing-extensions>=4.0.0