📖 Book Recommender System - Flask API

---

🔄 Full Process: From .NET Request to JSON Response

📁 1. .NET Client Sends a POST Request

The .NET frontend or backend sends a POST request to your Flask API endpoint:

POST http://localhost:5000/recommend
Content-Type: application/json

Example JSON body:

{
  "age": 25,
  "gender": "Male",
  "country": "Egypt",
  "is_new_muslim": "No",
  "born_muslim": "Yes",
  "education_level": "Bachelor",
  "religious_level": "Moderate",
  "preferred_topic": "Faith"
}

---

🧬 2. Flask API Receives the Request

Your Flask app has an endpoint defined like this:

@app.route("/recommend", methods=["POST"])
def recommend():
    data = request.json

The incoming JSON is parsed and stored as a Python dictionary.

---

🔮 3. Preprocess the Input

• The dictionary is converted into a DataFrame:

user_df = pd.DataFrame([data])

• Then it's encoded and scaled using saved preprocessing tools (e.g., joblib-encoded encoders and scalers):

user_df["gender"] = gender_encoder.transform(user_df["gender"])
user_df_scaled = scaler.transform(user_df[numerical_features])

---

🤖 4. Load the Model and Predict

• The saved model (e.g., TensorFlow/Keras) is loaded:

model = load_model("model.h5")
prediction = model.predict([user_inputs])

• The model returns predictions for all books.

---

📚 5. Select Top Recommended Books

• The predicted scores are sorted and top N indices are selected:

top_indices = prediction[0].argsort()[::-1][:5]
recommended_books = books_df.iloc[top_indices]

---

📤 6. Return Recommendations as JSON Response

• Flask sends the selected recommendations back to the .NET client as JSON:

return jsonify({
    "recommendations": recommended_books.to_dict(orient="records")
})

---

✅ Summary of Steps

Step	Description
1	Receive POST request from .NET
2	Parse incoming JSON
3	Preprocess data (encode/scale)
4	Feed data into trained model
5	Get top-N predictions
6	Return recommendations as JSON

---

🏦 Hosting

This API is typically run locally for testing:

python app.py

Visit: http://localhost:5000/recommend

Make sure the following files exist:

• model.h5 (trained model)
• encoder.pkl, scaler.pkl (preprocessing tools)
• books.csv or any book metadata file

---