adding app file

app.py

@@ -0,0 +1,110 @@
+import gradio as gr
+import pandas as pd
+import base64
+import os
+from openai import OpenAI
+
+client = OpenAI()
+
+def process_excel_file(file_path, non_question_columns_str):
+    # Load the Excel file into a DataFrame
+    df = pd.read_excel(file_path, header=0, index_col=None)
+
+    # Convert non-question columns string to a list and make it lowercase for comparison
+    non_question_columns = [col.strip().lower() for col in non_question_columns_str.split(',')]
+    # Identify columns that are considered as questions and non-questions
+    question_columns = [col for col in df.columns if col.lower() not in non_question_columns]
+    non_question_data = df[[col for col in df.columns if col.lower() in non_question_columns]]
+
+    # Filter the DataFrame to include only question columns
+    df_questions = df[question_columns]
+    # Drop Timestamp column
+    df_questions.drop(columns=["Timestamp"], inplace=True)
+    
+    
+    # Extract headers as questions
+    questions = list(df_questions.columns)
+
+    # Extract the last row as answer keys
+    answer_keys = df_questions.iloc[-1].to_dict()
+
+    # Extract participant answers (excluding the last row which is the answer keys)
+    participant_answers = df_questions.iloc[:-1]
+
+    return questions, answer_keys, participant_answers, non_question_data
+
+def process_and_grade(file_path, non_question_columns_str):
+    questions, answer_keys, participant_answers, non_question_data = process_excel_file(file_path, non_question_columns_str)
+
+    # Initialize columns for individual question scores based on numeric indices and total score
+    num_questions = len(questions)
+    for i in range(num_questions):
+        participant_answers[f"Q{i+1} Score"] = 0  # Creates score columns .
+    participant_answers['Total Points'] = 0
+    participant_answers['Grade'] = 0
+
+    # Grade each participant's answers and calculate total score
+    for index, row in participant_answers.iterrows():
+        total_score = 0
+        for i, question in enumerate(questions, start=1):
+            participant_answer = row[question]
+            correct_answer = answer_keys[question]
+            score = grade_answer(question, participant_answer, correct_answer)
+            participant_answers.at[index, f"Q{i} Score"] = score  # Update score using numeric index
+            total_score += score
+        participant_answers.at[index, 'Total Points'] = total_score  # Update total points
+        participant_answers.at[index, 'Grade'] = (total_score/(num_questions*2))*100 
+
+    # Concatenate non-question data with updated participant answers for final DataFrame
+    final_df = pd.concat([non_question_data.iloc[:-1].reset_index(drop=True), participant_answers.reset_index(drop=True)], axis=1)
+
+    # Save to a new Excel file
+    output_file_path = "graded_results.xlsx"
+    final_df.to_excel(output_file_path, index=False)
+
+    return output_file_path
+
+
+def grade_answer(question, participant_answer, key_answer):
+    """
+    Call the fine-tuned model to grade an answer based on the question,
+    the participant's answer, and the key answer.
+    """
+    if participant_answer.strip() == "-" or participant_answer.strip() == "":
+        return 0
+
+    task="Evaluate the correctness of a participant's answer compared to the intended answer"
+    prompt = f"{task}, Question: {question}\nIntended Answer: {key_answer}\nParticipant Answer: {participant_answer}\n"
+
+
+    response = client.completions.create(
+        model=os.environ['model_ft_id'],
+        prompt=prompt,
+        max_tokens=1,
+        temperature=0,
+    )
+
+    # Assuming the model's response is directly the score (0, 1, or 2)
+    try:
+        score = int(response.choices[0].text.strip())
+        return score
+    except ValueError:
+        # Handle unexpected model response
+        return 0
+
+
+# Set up the Gradio interface
+iface = gr.Interface(
+    fn=process_and_grade,
+    inputs=[
+        gr.File(file_count="single", type="filepath"),
+        gr.Textbox(lines=2, placeholder="Column1, Column2, Column3, ...", label="Non-Question Columns"),
+        # Remove model_id if it's set globally or passed in another way
+    ],
+    outputs=gr.File(label="Download Graded Results"),
+    title="Essay Question Grading",
+    description="Upload a spreadsheet and specify non-question columns (separated by commas) to grade. Format your Excel file with the 1st row for questions and the last row for key answers.",
+    allow_flagging="never"
+)
+
+iface.launch()