Spam-Mail-Classifier

An end to end ML Flow

Spam Mail Classifier & Monitoring

This repository contains a complete end-to-end solution for training, serving, and monitoring a spam-mail classifier using:

• Python (data preparation & model training)
• Flask (REST API with /predict & /metrics endpoints)
• Docker (containerization)
• Kubernetes (Deployment, Service, CronJob)
• Prometheus (scraping custom & system metrics)
• Grafana (dashboarding)

---

Repository Structure

📦 Spam-Mail-Classifier
├── data/                      # raw & processed datasets
│   └── ...
├── processed_data/            # train/val/test splits, tokenizer
├── best_model.keras           # trained Keras model
├── tokenizer.pickle           # fitted tokenizer
├── app.py                     # Flask application
├── Train.py                   # data prep & training CLI
├── Dockerfile                 # Docker image spec
├── deployment.yaml            # K8s Deployment & Service for Spam API
├── service.yaml               # (alternative) K8s Service manifest
├── retrain-cronjob.yaml       # K8s CronJob to retrain the model periodically
├── extra-scrape-configs.yaml  # Prometheus scrape configs for Kubernetes
├── Grafana-deployment.yaml    # Grafana Deployment & Service in K8s
├── requirements.txt           # Python dependencies
└── README.md                  # this file

---

Quickstart

1. Clone & Setup

git clone https://github.com/your-org/spam-mail-classifier.git
cd spam-mail-classifier

Create a Python virtual environment and install dependencies:

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Prepare Data & Train Model

python Train.py \
  --csv spam.csv \
  --output-dir data \
  --num-words 10000 \
  --maxlen 100 \
  --test-size 0.2 \
  --val-size 0.1 \
  --seed 42

3. Build & Push Docker Image

docker build -t <dockerhub-username>/spam-api:latest .
docker push <dockerhub-username>/spam-api:latest

4. Deploy to Kubernetes

Make sure Docker Desktop’s Kubernetes is enabled (or point kubectl to your cluster).

kubectl apply -f deployment.yaml
kubectl apply -f service.yaml

Check pods and service:

kubectl get pods -l app=spam-api
kubectl get svc spam-api-service

5. Test the API

Port-forward or use the NodePort:

# Port-forward locally:
kubectl port-forward svc/spam-api-service 5000:80
curl -X POST http://localhost:5000/predict \
  -H "Content-Type: application/json" \
  -d '{"text":"Free tickets waiting for you"}'

6. Monitoring with Prometheus & Grafana

1. Prometheus

   helm install prometheus prometheus-community/prometheus \
     --set server.persistentVolume.enabled=false
   kubectl port-forward svc/prometheus-server 9090:80

2. Grafana

   helm install grafana grafana/grafana \
     --set service.type=NodePort \
     --set service.nodePort=30081 \
     --set adminPassword="letmein"
   kubectl port-forward svc/grafana 30082:80

   • Access Grafana at http://localhost:30082
   • Add Prometheus (URL: http://prometheus-server.monitoring.svc.cluster.local:80)
   • Import community dashboards (e.g. Node Exporter Full)
   • Build custom panels for spamapi_predictions_total

---

Next Steps & Customization

• Autoscaling: Add a HorizontalPodAutoscaler on CPU or custom Prometheus metrics.
• CI/CD: Automate image builds & kubectl apply with GitHub Actions.
• Alerting: Define Prometheus alerting rules and hook Alertmanager.
• Persistence: Enable persistent volumes for Prometheus & Grafana.
• Ingress: Use an ingress controller (e.g., NGINX) for external access.

---

Happy monitoring and spam‑fighting!