BookFinder

BookFinder is a web application that helps users discover books similar to their favorites by analyzing book descriptions and genres. It leverages natural language processing to compare descriptions and generate personalized recommendations from a custom dataset.

---

Features

• 🔍 Search by title & author using the Google Books API
• ✂️ Summarize descriptions with a pretrained t5-small model
• 📚 Find similar books using sentence embeddings
• 🔗 Clickable Libgen links for quick access to recommended titles

---

Tech Stack

• Frontend: Streamlit
• Backend: Flask API + precomputed embeddings
• ML Models: Transformers (T5), Sentence-Transformers
• Data Handling: Pandas, NumPy
• Infrastructure: Docker, Kubernetes, Helm, Kustomize, Terraform, GKE Autopilot
• APIs: Google Books API

---

Architecture Overview

• Frontend (Streamlit) runs on port 8501, calls backend API
• Backend (Flask) runs on port 5000, loads a CSV with 500+ books and precomputed embeddings
• Deployment: Dockerized, deployed via Kubernetes
• Cloud Option: Full deployment via Terraform to GCP using GKE Autopilot

---

Repository Structure

BookFinder/
├── app/                # Frontend (Streamlit)
│   ├── app.py
│   ├── Dockerfile
│   └── requirements.txt
├── backend/            # Backend (Flask)
│   ├── app.py
│   ├── book_embeddings.csv
│   ├── Dockerfile
│   └── requirements.txt
├── deployments/        # Deployment configuration
│   ├── helm/           # Helm charts for Kubernetes deployment
│   │   └── bookfinder/ # Main chart directory
│   │       ├── Chart.yaml
│   │       ├── values.yaml
│   │       └── templates/
│   │           ├── _helpers.tpl
│   │           ├── backend-deployment.yaml
│   │           ├── backend-service.yaml
│   │           ├── configmap.yaml
│   │           ├── frontend-deployment.yaml
│   │           ├── frontend-service.yaml
│   │           ├── ingress.yaml
│   │           └── secret.yaml
│   └── kustomize/      # Kustomize-ready Kubernetes manifests
│       ├── kustomization.yaml
│       ├── frontend-deployment.yaml
│       ├── frontend-service.yaml
│       ├── backend-deployment.yaml
│       ├── backend-service.yaml
│       ├── configmap.yaml
│       ├── ingress.yaml
│       └── secret.yaml
├── iac/                # Terraform files for GCP Autopilot
│   ├── main.tf
│   ├── variables.tf
│   └── terraform.tfvars
├── scripts/            # Utility scripts for dataset preparation and model training
│   └── generate_embeddings.py
└── README.md

---

🔧 Deployment Options

Local Development (Kustomize)

For local development and testing, I recommend using Kustomize with minikube:

# 1. Clone the repository
git clone https://github.com/bennymestel/BookFinder.git
cd BookFinder

# 2. Start minikube
minikube start

# 3. Add a local DNS entry
# For macOS/Linux:
echo "127.0.0.1 bookfinder.local" | sudo tee -a /etc/hosts
# For Windows (run PowerShell as Administrator):
Add-Content -Path "$env:windir\System32\drivers\etc\hosts" -Value "`n127.0.0.1 bookfinder.local" -Force

# 4. Apply manifests using Kustomize
kubectl apply -k deployments/kustomize/

# 5. Start a tunnel for ingress access (keep this terminal open)
minikube tunnel

Once deployed, BookFinder will be available at: http://bookfinder.local

You can check the deployment status with:

kubectl get pods
kubectl get ingress

Note: Both frontend and backend services may take a few minutes to become fully operational on initial startup. The frontend needs time to download and initialize ML models (T5 and Sentence-Transformers), while the backend loads embeddings and prepares the recommendation system.

GCP Cloud Deployment (Helm)

For GCP deployments, I recommend using Helm which is pre-configured for GCP's load balancing:

---

☁️ Cloud Deployment (GCP Autopilot via Terraform)

Prerequisites

• Google Cloud SDK
• Terraform
• A GCP account with billing enabled
• Docker images published on DockerHub:
  • bennymestel/book-finder-frontend
  • bennymestel/book-finder-backend

Steps

1. Set up GKE Autopilot with Terraform:

# Clone and enter the Terraform directory
cd BookFinder/iac

# Set up variables
# Edit terraform.tfvars with your GCP project ID and region

# Authenticate with GCP
gcloud auth login
gcloud auth application-default login

# Deploy the infrastructure
terraform init
terraform apply

# Connect to GKE (replace REGION with the region you specified in terraform.tfvars)
gcloud container clusters get-credentials book-finder-cluster --region=REGION

2. Deploy BookFinder using Helm:

# Deploy using Helm with GCP-optimized settings
helm install bookfinder ../deployments/helm/bookfinder

# Get external IP (may take 5-10 minutes to provision)
kubectl get ingress

Access your deployed app at the external IP address shown:

http://<EXTERNAL-IP>

Note: Both frontend and backend services may take a few minutes to become fully operational on initial startup. The frontend needs time to download and initialize ML models (T5 and Sentence-Transformers), while the backend loads embeddings and prepares the recommendation system.

---

🔐 Secrets and API Keys

Store your Google Books API key securely:

• In deployments/kustomize/secret.yaml or deployments/helm/bookfinder/templates/secret.yaml
• Or as an environment variable

Streamlit config and secrets can be set via mounted secrets.toml.

---

🔁 Updating Embeddings

To regenerate book_embeddings.csv:

python scripts/generate_embeddings.py

Replace the CSV in backend/, rebuild the image, and redeploy.

---

📦 Docker Images

Prebuilt images are hosted on DockerHub:

• Frontend: bennymestel/book-finder-frontend
• Backend: bennymestel/book-finder-backend

---

✅ Status

• App runs locally and in the cloud
• Fully containerized and K8s-deployable
• Multiple deployment options: Kustomize and Helm
• Infrastructure as Code with Terraform
• Public GCP-compatible deployment path
• Kubernetes Ingress for unified access