main.py

import os
import tempfile
import io
import json
import numpy as np
import cv2
from PIL import Image
from pdf2image import convert_from_bytes
from fastapi import FastAPI, UploadFile, File, HTTPException
from fastapi.responses import JSONResponse, StreamingResponse
import uvicorn

# Get API key from environment
GENAI_API_KEY = os.getenv("GENAI_API_KEY")
if not GENAI_API_KEY:
    raise Exception("GENAI_API_KEY not set in environment")

# Import the Google GenAI client libraries.
from google import genai
from google.genai import types

# Initialize the GenAI client with the API key.
client = genai.Client(api_key=GENAI_API_KEY)

app = FastAPI(title="Student Result Card API")

# Use system temporary directory to store the results file.
TEMP_FOLDER = tempfile.gettempdir()
RESULT_FILE = os.path.join(TEMP_FOLDER, "result_cards.json")

##############################################################
# Preprocessing & Extraction Functions
##############################################################

def extract_json_from_output(output_str: str):
    """
    Extracts a JSON object from a string containing extra text.
    """
    start = output_str.find('{')
    end = output_str.rfind('}')
    if start == -1 or end == -1:
        print("No JSON block found in the output.")
        return None
    json_str = output_str[start:end+1]
    try:
        result = json.loads(json_str)
        return result
    except json.JSONDecodeError as e:
        print("Error decoding JSON:", e)
        return None

def parse_all_answers(image_input: Image.Image) -> str:
    """
    Extracts answers from an image of a 15-question answer sheet.
    Returns the response text (JSON string).
    """
    output_format = """
Answer in the following JSON format. Do not write anything else:
{
  "Answers": {
    "1": "<option or text>",
    "2": "<option or text>",
    "3": "<option or text>",
    "4": "<option or text>",
    "5": "<option or text>",
    "6": "<option or text>",
    "7": "<option or text>",
    "8": "<option or text>",
    "9": "<option or text>",
    "10": "<option or text>",
    "11": "<free-text answer>",
    "12": "<free-text answer>",
    "13": "<free-text answer>",
    "14": "<free-text answer>",
    "15": "<free-text answer>"
  }
}
"""
    prompt = f"""
You are an assistant that extracts answers from an image.
The image is a screenshot of an answer sheet containing 15 questions.
For questions 1 to 10, the answers are multiple-choice selections.
For questions 11 to 15, the answers are free-text responses.
Extract the answer for each question (1 to 15) and provide the result in JSON using the format below:
{output_format}
"""
    response = client.models.generate_content(
        model="gemini-2.0-flash",
        contents=[prompt, image_input]
    )
    return response.text

def parse_info(image_input: Image.Image) -> str:
    """
    Extracts candidate information including name, number, country, level and paper from an image.
    Returns the response text (JSON string).
    """
    output_format = """
Answer in the following JSON format. Do not write anything else:
{
  "Candidate Info": {
    "Name": "<name>",
    "Number": "<number>",
    "Country": "<country>",
    "Level": "<level>",
    "Paper": "<paper>"
  }
}
"""
    prompt = f"""
You are an assistant that extracts candidate information from an image.
The image contains candidate details including name, candidate number, country, level and paper.
Extract the information accurately and provide the result in JSON using the following format:
{output_format}
"""
    response = client.models.generate_content(
        model="gemini-2.0-flash",
        contents=[prompt, image_input]
    )
    return response.text

def parse_paper(student_info_text: str) -> str:
    """
    Extracts the Paper field from candidate information.
    Returns the paper letter (e.g. "A", "B", or "K") as a string.
    """
    prompt = f"""
You are an assistant that extracts the Paper from candidate information.
The candidate information contains details including their paper designation.
Extract the Paper value (one alphabet only) from the following:
{student_info_text}
"""
    response = client.models.generate_content(
        model="gemini-2.0-flash",
        contents=[prompt, student_info_text]
    )
    return response.text.strip()

def calculate_result(student_answers: dict, correct_answers: dict) -> dict:
    """
    Compares student's answers with the correct answers and calculates the score.
    Assumes JSON structures with a top-level "Answers" key containing Q1 to Q15.
    """
    student_all = student_answers.get("Answers", {})
    correct_all = correct_answers.get("Answers", {})
    total_questions = 15
    marks = 0
    detailed = {}
    
    for q in map(str, range(1, total_questions + 1)):
        stud_ans = student_all.get(q, "").strip()
        corr_ans = correct_all.get(q, "").strip()
        if stud_ans == corr_ans:
            marks += 1
            detailed[q] = {"Student": stud_ans, "Correct": corr_ans, "Result": "Correct"}
        else:
            detailed[q] = {"Student": stud_ans, "Correct": corr_ans, "Result": "Incorrect"}
    
    percentage = (marks / total_questions) * 100
    result_card = {
        "Total Marks": marks,
        "Total Questions": total_questions,
        "Percentage": percentage,
        "Detailed Results": detailed
    }
    return result_card

##############################################################
# Helper: Load and Process an Answer Key PDF (from bytes)
##############################################################
def load_answer_key(pdf_bytes: bytes) -> dict:
    """
    Converts a PDF (as bytes) to images, extracts the last page, and parses the answers.
    Returns the parsed JSON answer key.
    """
    images = convert_from_bytes(pdf_bytes)
    last_page_image = images[-1]
    answer_key_response = parse_all_answers(last_page_image)
    answer_key = extract_json_from_output(answer_key_response)
    return answer_key

##############################################################
# FastAPI Endpoints
##############################################################

@app.post("/process")
async def process_pdfs(
    student_pdf: UploadFile = File(..., description="PDF with all student answer sheets (one page per student)"),
    paper_a_pdf: UploadFile = File(..., description="Answer key PDF for Paper A"),
    paper_b_pdf: UploadFile = File(..., description="Answer key PDF for Paper B"),
    paper_k_pdf: UploadFile = File(..., description="Answer key PDF for Paper K")
):
    try:
        # Read file bytes
        student_pdf_bytes = await student_pdf.read()
        paper_a_bytes = await paper_a_pdf.read()
        paper_b_bytes = await paper_b_pdf.read()
        paper_k_bytes = await paper_k_pdf.read()
        
        # Preload answer keys from the three PDFs
        answer_keys = {
            "A": load_answer_key(paper_a_bytes),
            "B": load_answer_key(paper_b_bytes),
            "K": load_answer_key(paper_k_bytes)
        }
        
        # Convert the student answer PDF to images (each page = one student)
        student_images = convert_from_bytes(student_pdf_bytes)
        all_results = []
        
        # Loop over all student pages
        for idx, page in enumerate(student_images):
            print(f"Processing student page {idx+1}...")
            
            # Convert the PIL image to OpenCV format for masking
            page_cv = np.array(page)
            page_cv = cv2.cvtColor(page_cv, cv2.COLOR_RGB2BGR)
            height, width = page_cv.shape[:2]
            
            ###########################################################
            # 1. Extract Candidate Information Region
            ###########################################################
            candidate_mask = np.zeros((height, width), dtype="uint8")
            candidate_margin_top = int(height * 0.10)
            candidate_margin_bottom = int(height * 0.75)
            cv2.rectangle(candidate_mask, (0, candidate_margin_top), (width, height - candidate_margin_bottom), 255, -1)
            masked_candidate = cv2.bitwise_and(page_cv, page_cv, mask=candidate_mask)
            coords = cv2.findNonZero(candidate_mask)
            if coords is None:
                continue  # Skip page if no candidate region is found.
            x, y, w, h = cv2.boundingRect(coords)
            cropped_candidate = masked_candidate[y:y+h, x:x+w]
            candidate_pil = Image.fromarray(cv2.cvtColor(cropped_candidate, cv2.COLOR_BGR2RGB))
            
            # Extract candidate info using GenAI.
            candidate_info_response = parse_info(candidate_pil)
            candidate_info = extract_json_from_output(candidate_info_response)
            
            # Determine the candidate's paper.
            paper = ""
            if candidate_info and "Candidate Info" in candidate_info:
                paper = candidate_info["Candidate Info"].get("Paper", "").strip()
            if not paper:
                paper = parse_paper(candidate_info_response)
            paper = paper.upper()
            print(f"Student {idx+1} Paper: {paper}")
            
            # Retrieve the appropriate answer key.
            if paper not in answer_keys or answer_keys[paper] is None:
                print(f"Error: Invalid or missing answer key for paper '{paper}' for student {idx+1}. Skipping.")
                continue
            correct_answer_key = answer_keys[paper]
            
            ###########################################################
            # 2. Extract Student Answers from the Entire Page
            ###########################################################
            student_answers_response = parse_all_answers(page)
            student_answers = extract_json_from_output(student_answers_response)
            
            ###########################################################
            # 3. Calculate the Result for this Student
            ###########################################################
            result = calculate_result(student_answers, correct_answer_key)
            
            # Compile the result for this student.
            result_card = {
                "Student Index": idx + 1,
                "Candidate Info": candidate_info.get("Candidate Info", {}) if candidate_info else {},
                "Student Answers": student_answers,
                "Correct Answer Key": correct_answer_key,
                "Result": result
            }
            all_results.append(result_card)
        
        # Write the results to a file in the temporary folder.
        with open(RESULT_FILE, "w", encoding="utf-8") as f:
            json.dump({"results": all_results}, f, indent=2)
        
        return JSONResponse(content={"results": all_results})
    
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))

@app.get("/download")
async def download_results():
    """
    Returns the result JSON file stored in the temporary folder.
    """
    if not os.path.exists(RESULT_FILE):
        raise HTTPException(status_code=404, detail="Result file not found. Please run /process first.")
    return StreamingResponse(
        open(RESULT_FILE, "rb"),
        media_type="application/json",
        headers={"Content-Disposition": f"attachment; filename=result_cards.json"}
    )

@app.get("/")
async def root():
    return {
        "message": "Welcome to the Student Result Card API.",
        "usage": "POST PDFs to /process (student answer sheet, paper A, paper B, paper K). Then use /download to retrieve the results."
    }

if __name__ == "__main__":
    uvicorn.run("main:app", host="0.0.0.0", port=8000, reload=True)