handler.py

from typing import Dict, List, Any
import pickle
import os
import __main__
import numpy as np

class CollaborativeRecommender:
    def __init__(self, algo, trainset):
        self.algo = algo
        self.trainset = trainset

    def predict(self, user_id, k=10):
        try:
            # Convert raw user_id to inner user_id
            inner_user_id = self.trainset.to_inner_uid(user_id)
        except ValueError:
            # User not found in trainset, return None
            return None

        # Get the list of books the user has interacted with
        user_books = set(self.trainset.ur[inner_user_id])
        all_books = set(self.trainset.all_items())
        unseen_books = all_books - user_books
        # Predict the ratings for unseen books
        predictions = [self.algo.predict(self.trainset.to_raw_uid(inner_user_id), self.trainset.to_raw_iid(book_id)) for book_id in unseen_books]
        # Sort the predictions by estimated rating and return the top-k books
        top_predictions = sorted(predictions, key=lambda x: x.est, reverse=True)[:k]
        top_books = [pred.iid for pred in top_predictions]
        return top_books

__main__.CollaborativeRecommender = CollaborativeRecommender

class EndpointHandler:
    def __init__(self, path=""):
        model_path = os.path.join(path, "model.pkl")
        with open(model_path, 'rb') as f:
            self.model = pickle.load(f)

    def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
        # Extract the 'inputs' from the data
        inputs = data.get('inputs', {})
        
        # If inputs is a string (for single user_id input), convert it to a dict
        if isinstance(inputs, str):
            inputs = {'user_id': inputs}
        
        user_id = inputs.get('user_id')
        k = inputs.get('k', 10)  # Default to 10 if not provided

        if user_id is None:
            return [{"error": "user_id is required"}]

        try:
            recommended_books = self.model.predict(user_id, k=k)
            return [{"recommended_books": recommended_books}]
        except Exception as e:
            return [{"error": str(e)}]

def load_model(model_path):
    handler = EndpointHandler(model_path)
    return handler