Create app.py

app.py

@@ -0,0 +1,83 @@
+from flask import Flask, request, jsonify
+from flask_cors import CORS
+from transformers import pipeline
+import nltk
+import ssl
+
+# This is a workaround for an NLTK download issue that can occur on some systems.
+try:
+    _create_unverified_https_context = ssl._create_unverified_context
+except AttributeError:
+    pass
+else:
+    ssl._create_default_https_context = _create_unverified_https_context
+
+# Download the 'punkt' tokenizer for splitting text into sentences
+nltk.download('punkt')
+
+# --- Model Loading ---
+# We use a T5 (Transformer) model fine-tuned for Question Generation.
+# T5 is a modern and powerful alternative to BERT for this specific task.
+# The model 'valhalla/t5-base-qg-hl' is designed to generate questions from text.
+print("Loading question generation model...")
+# Using a specific revision to ensure consistency
+question_generator = pipeline("text2text-generation", model="valhalla/t5-base-qg-hl")
+print("Model loaded successfully!")
+
+
+# --- Flask App Setup ---
+app = Flask(__name__)
+# Enable Cross-Origin Resource Sharing (CORS) to allow your frontend
+# to communicate with this API.
+CORS(app)
+
+# --- API Endpoint ---
+@app.route('/generate-quiz', methods=['POST'])
+def generate_quiz():
+    """
+    This function handles the API request to generate a quiz.
+    It expects a JSON payload with a 'context' key.
+    """
+    # Get the JSON data from the request
+    data = request.get_json()
+    if not data or 'context' not in data:
+        return jsonify({'error': 'No context provided in the request.'}), 400
+
+    context = data['context']
+
+    # 1. Split the context into individual sentences.
+    sentences = nltk.sent_tokenize(context)
+
+    # We'll limit the number of questions to avoid long processing times.
+    # You can adjust this limit as needed.
+    max_questions = 10
+    sentences = sentences[:max_questions]
+
+    generated_questions = []
+
+    print(f"Generating questions for {len(sentences)} sentences...")
+
+    # 2. Generate a question for each sentence.
+    for sentence in sentences:
+        try:
+            # The model generates a string which might contain multiple questions
+            # separated by <sep>. We take the first one.
+            output = question_generator(sentence, max_length=64, num_beams=4)
+            question = output[0]['generated_text'].split('<sep>')[0].strip()
+            # Basic validation to ensure the question is meaningful
+            if question.endswith('?') and len(question.split()) > 3:
+                 generated_questions.append({'question': question, 'context': sentence})
+        except Exception as e:
+            print(f"Error generating question for sentence: '{sentence}'. Error: {e}")
+
+
+    print(f"Successfully generated {len(generated_questions)} questions.")
+
+    # 3. Return the list of questions as a JSON response.
+    return jsonify({'quiz': generated_questions})
+
+# --- Run the App ---
+if __name__ == '__main__':
+    # When running locally, the app will be available at http://127.0.0.1:5000
+    # When running in Colab, you will need a tool like ngrok to expose it.
+    app.run(host='0.0.0.0', port=5000)