Update app.py

app.py

@@ -1,10 +1,34 @@
 import gradio as gr
-import PyPDF2
-import json
-from some_llm_api import call_gemini_llm  # replace with actual API call
+import os
+from PIL import Image
+import google.generativeai as genai
 
-# Predefined topic list (abbreviated for brevity; include all topics in practice)
-TOPICS = [
+# Configure Gemini API
+genai.configure(api_key=os.getenv("GEMINI_API_KEY"))
+
+def create_model():
+    try:
+        return genai.GenerativeModel("gemini-2.5-pro", generation_config={"temperature": 0})
+    except:
+        return genai.GenerativeModel("gemini-2.5-flash", generation_config={"temperature": 0})
+
+def gemini_generate_content(model, prompt_text, file_upload_obj=None):
+    inputs = [prompt_text]
+    if file_upload_obj:
+        inputs.append(file_upload_obj)
+    response = model.generate_content(inputs)
+    raw_text = getattr(response, "text", None)
+    if not raw_text and getattr(response, "candidates", None):
+        raw_text = response.candidates[0].content.parts[0].text
+    if not raw_text:
+        raw_text = str(response)
+    return raw_text
+
+# Load merged_gemini_output.txt reference
+with open("merged_gemini_output.txt", "r", encoding="utf-8") as f:
+    merged_questions_ref = f.read()
+
+TOPIC_LIST = """
     "SL 1.1 - Operations with numbers in the form a × 10k where 1 < a < 10 and k is an integer.",
     "SL 1.2 - Arithmetic sequences and series. Use of the formulae for the nth term and the sum of the first n terms of the sequence. Use of sigma notation for sums of arithmetic sequences. Applications. Analysis, interpretation and prediction where a model is not perfectly arithmetic in real life.",
     "SL 1.3 - Geometric sequences and series. Use of the formulae for the n th term and the sum of the first n terms of the sequence. Use of sigma notation for the sums of geometric sequences. Applications.",
@@ -89,72 +113,56 @@ TOPICS = [
     "AHL 5.18 - First order differential equations. Numerical solution of dy/dx = f(x, y) using Euler's method. Variables separable. Homogeneous differential equation. Solution of y' + P(x)y = Q(x), using the integrating factor.",
     "AHL 5.19 - Maclaurin series to obtain expansions for eˣ, sinx, cosx, ln(1+x), (1+x)ᵖ, p∈Q. Use of simple substitution, products, integration and differentiation to obtain other series. Maclaurin series developed from differential equations"
 
+"""
+
+def identify_topics_and_generate_test(qp_file_obj, graded_as_file_obj):
+    model = create_model()
+    qp_path = qp_file_obj.name
+    as_path = graded_as_file_obj.name
+    
+    prompt = f"""
+You are an experienced math examiner and curriculum expert.
+
+INPUT:
+- Question Paper PDF: {qp_path}
+- Graded Answer Sheet PDF: {as_path}
+
+TASK 1: Identify the topic of each question in the Question Paper from this list:
+{TOPIC_LIST}
+
+Refer to previous questions and topics provided here:
+{merged_questions_ref}
+
+OUTPUT 1: Provide a JSON array of questions in the following format:
+[
+  {{"question_number": 1, "topic": "SL 1.2", "content": "..."}},
+  ...
 ]
 
-def extract_pdf_text(pdf_file):
-    reader = PyPDF2.PdfReader(pdf_file)
-    text = ""
-    for page in reader.pages:
-        text += page.extract_text() + "\n"
-    return text
-
-def process_qp_and_graded(qp_file, graded_file):
-    # Step 1: Extract text
-    qp_text = extract_pdf_text(qp_file)
-    graded_text = extract_pdf_text(graded_file)
-
-    # Step 2: Call Gemini LLM to identify topics for each question
-    llm_prompt = f"""
-    You are a math expert. Identify the topic for each question in the following question paper
-    from this list: {', '.join(TOPICS)}. 
-    Return JSON in the following format:
-    [
-      {{
-        "question_number": 1,
-        "topic": "SL 1.1",
-        "content": "The question text"
-      }},
-      ...
-    ]
-    Question paper text:
-    {qp_text}
-    """
-    identified_questions = call_gemini_llm(llm_prompt)
-
-    # Step 3: Generate new 8-question test based on student's graded answers
-    # Prompt LLM to generate new test using the recommended composition
-    llm_test_prompt = f"""
-    You are a math teacher. Based on the graded answers below:
-    {graded_text}
-
-    And the following questions with topics:
-    {identified_questions}
-
-    Create a new 8-question test with the following composition:
-    - 4 remediation items based on wrong answers (2 near-transfer, 2 far-transfer)
-    - 2 retention items on topics the student got right
-    - 1 synthesis / higher-order item combining multiple concepts
-    - 1 confidence/metacognition item
-
-    Return JSON with question_number, topic, and content.
-    """
-    new_test = call_gemini_llm(llm_test_prompt)
-
-    return json.dumps(identified_questions, indent=2), json.dumps(new_test, indent=2)
-
-# Gradio interface
-iface = gr.Interface(
-    fn=process_qp_and_graded,
-    inputs=[
-        gr.File(label="Question Paper PDF"),
-        gr.File(label="Graded Answers PDF")
-    ],
-    outputs=[
-        gr.Textbox(label="Questions with Topics", lines=20),
-        gr.Textbox(label="Generated 8-Question Test", lines=20)
-    ],
-    title="Math Question Topic Identifier & Test Generator",
-    description="Upload the student's question paper and graded answers PDFs. This app identifies question topics and generates a new targeted test."
-)
-
-iface.launch()
+TASK 2: Based on the student's performance in the graded answer sheet, generate a new **8-question remedial test** following these rules:
+1. 4 remediation items: concepts the student got wrong (2 near-transfer, 2 far-transfer)
+2. 2 retention items: concepts the student got right
+3. 1 synthesis/higher-order item: combines multiple concepts
+4. 1 confidence/metacognition item: student reflects on confidence or explains an answer
+
+OUTPUT 2: Provide a JSON array of the 8 remedial questions with question number, topic, and content.
+"""
+    
+    result = gemini_generate_content(model, prompt_text=prompt)
+    return result
+
+with gr.Blocks(title="AI Topic Identifier & Remedial Test Generator") as demo:
+    gr.Markdown("## 📘 Topic Identification & Remedial Test Generator")
+
+    with gr.Row():
+        qp_file = gr.File(label="📄 Upload Question Paper (PDF)")
+        graded_as_file = gr.File(label="📝 Upload Graded Answer Sheet (PDF)")
+
+    run_button = gr.Button("🚀 Identify Topics & Generate Test")
+
+    output_box = gr.Textbox(label="📝 Gemini Output (JSON)", lines=20)
+
+    run_button.click(fn=identify_topics_and_generate_test, inputs=[qp_file, graded_as_file], outputs=[output_box])
+
+if __name__ == "__main__":
+    demo.launch()