# CHARTS + DOWNLOAD + NO NAMES # intervention_analysis_app.py #------------------------------------------------------------------------ # Import Modules #------------------------------------------------------------------------ import streamlit as st import pandas as pd import matplotlib.pyplot as plt import io import re # from transformers import pipeline from huggingface_hub import InferenceClient import os from pathlib import Path from dotenv import load_dotenv load_dotenv() #------------------------------------------------------------------------ # Configurations #------------------------------------------------------------------------ # Streamlit page setup st.set_page_config( page_title="Intervention Program Analysis", page_icon=":bar_chart:", layout="centered", initial_sidebar_state="auto", menu_items={ 'Get Help': 'mailto:clevesseur@mimtss.org', 'About': "This app is built to support spreadsheet analysis" } ) #------------------------------------------------------------------------ # Sidebar #------------------------------------------------------------------------ with st.sidebar: # Password input field # password = st.text_input("Enter Password:", type="password") # Set the desired width in pixels image_width = 300 # Define the path to the image image_path = "mimtss.png" # Display the image st.image(image_path, width=image_width) # Toggle for Help and Report a Bug with st.expander("Need help and report a bug"): st.write(""" **Contact**: Cheyne LeVesseur, PhD **Email**: clevesseur@mimtss.org """) st.divider() st.subheader('User Instructions') # Principles text with Markdown formatting User_Instructions = """ - **Step 1**: Upload your Excel file. - **Step 2**: Anonymization – student names are replaced with initials for privacy. - **Step 3**: Review anonymized data. - **Step 4**: View **intervention session statistics**. - **Step 5**: Review **student attendance and engagement metrics**. - **Step 6**: Review AI-generated **insights and recommendations**. ### **Privacy Assurance** - **No full names** are ever displayed or sent to the AI model—only initials are used. - This ensures that sensitive data remains protected throughout the entire process. ### **Detailed Instructions** #### **1. Upload Your Excel File** - Start by uploading an Excel file that contains intervention data. - Click on the **“Upload your Excel file”** button and select your `.xlsx` file from your computer. **Note**: Your file should have columns like "Did the intervention happen today?" and "Student Attendance [FirstName LastName]" for the analysis to work correctly. #### **2. Automated Name Anonymization** - Once the file is uploaded, the app will **automatically replace student names with initials** in the "Student Attendance" columns. - For example, **"Student Attendance [Cheyne LeVesseur]"** will be displayed as **"Student Attendance [CL]"**. - If the student only has a first name, like **"Student Attendance [Cheyne]"**, it will be displayed as **"Student Attendance [C]"**. - This anonymization helps to **protect student privacy**, ensuring that full names are not visible or sent to the AI language model. #### **3. Review the Uploaded Data** - You will see the entire table of anonymized data to verify that the information has been uploaded correctly and that names have been replaced with initials. #### **4. Intervention Session Statistics** - The app will calculate and display statistics related to intervention sessions, such as: - **Total Number of Days Available** - **Intervention Sessions Held** - **Intervention Sessions Not Held** - **Intervention Frequency (%)** - A **stacked bar chart** will be shown to visualize the number of sessions held versus not held. - If you need to save the visualization, click the **“Download Chart”** button to download it as a `.png` file. #### **5. Student Metrics Analysis** - The app will also calculate metrics for each student: - **Attendance (%)** – The percentage of intervention sessions attended. - **Engagement (%)** – The level of engagement during attended sessions. - These metrics will be presented in a **line graph** that shows attendance and engagement for each student. - You can click the **“Download Chart”** button to download the visualization as a `.png` file. #### **6. Generate AI Analysis and Recommendations** - The app will prepare data from the student metrics to provide notes, key takeaways, and suggestions for improving outcomes using an **AI language model**. - You will see a **spinner** labeled **“Generating AI analysis…”** while the AI processes the data. - This step may take a little longer, but the spinner ensures you know that the system is working. - Once the analysis is complete, the AI's recommendations will be displayed under **"AI Analysis"**. - You can click the **“Download LLM Output”** button to download the AI-generated recommendations as a `.txt` file for future reference. """ st.markdown(User_Instructions) #------------------------------------------------------------------------ # Functions #------------------------------------------------------------------------ # Set the Hugging Face API key # Retrieve Hugging Face API key from environment variables hf_api_key = os.getenv('HF_API_KEY') if not hf_api_key: raise ValueError("HF_API_KEY not set in environment variables") # Create the Hugging Face inference client client = InferenceClient(api_key=hf_api_key) # Constants INTERVENTION_COLUMN = 'Did the intervention happen today?' ENGAGED_STR = 'Engaged (Respect, Responsibility, Effort)' PARTIALLY_ENGAGED_STR = 'Partially Engaged (about 50%)' NOT_ENGAGED_STR = 'Not Engaged (less than 50%)' def main(): st.title("Intervention Program Analysis") # File uploader uploaded_file = st.file_uploader("Upload your Excel file", type=["xlsx"]) if uploaded_file is not None: try: # Read the Excel file into a DataFrame df = pd.read_excel(uploaded_file) # Format "Date of Session" and "Timestamp" df['Date of Session'] = pd.to_datetime(df['Date of Session']).dt.strftime('%m/%d/%Y') df['Timestamp'] = pd.to_datetime(df['Timestamp']).dt.strftime('%I:%M %p') df = df[['Date of Session', 'Timestamp'] + [col for col in df.columns if col not in ['Date of Session', 'Timestamp']]] # Replace student names with initials df = replace_student_names_with_initials(df) st.subheader("Uploaded Data") st.write(df) # Ensure expected column is available if INTERVENTION_COLUMN not in df.columns: st.error(f"Expected column '{INTERVENTION_COLUMN}' not found.") return # Clean up column names df.columns = df.columns.str.strip() # Compute Intervention Session Statistics intervention_stats = compute_intervention_statistics(df) st.subheader("Intervention Session Statistics") st.write(intervention_stats) # Two-column layout for the visualization and intervention frequency col1, col2 = st.columns(2) with col1: intervention_fig = plot_intervention_statistics(intervention_stats) with col2: intervention_frequency = intervention_stats['Intervention Frequency (%)'].values[0] st.markdown(f"

{intervention_frequency}%

", unsafe_allow_html=True) # Add download button for Intervention Session Statistics chart download_chart(intervention_fig, "intervention_statistics_chart.png") # Compute Student Metrics student_metrics_df = compute_student_metrics(df) st.subheader("Student Metrics") st.write(student_metrics_df) # Visualization for Student Metrics student_metrics_fig = plot_student_metrics(student_metrics_df) # Add download button for Student Metrics chart download_chart(student_metrics_fig, "student_metrics_chart.png") # Prepare input for the language model llm_input = prepare_llm_input(student_metrics_df) # Generate Notes and Recommendations using Hugging Face LLM with st.spinner("Generating AI analysis..."): recommendations = prompt_response_from_hf_llm(llm_input) st.subheader("AI Analysis") st.markdown(recommendations) # Add download button for LLM output download_llm_output(recommendations, "llm_output.txt") except Exception as e: st.error(f"Error reading the file: {str(e)}") def replace_student_names_with_initials(df): """Replace student names in column headers with initials.""" updated_columns = [] for col in df.columns: if col.startswith('Student Attendance'): # Extract the name from the column header match = re.match(r'Student Attendance \[(.+?)\]', col) if match: name = match.group(1) # Split the name into parts (first and last name) name_parts = name.split() # Convert the name to initials if len(name_parts) == 1: initials = name_parts[0][0] # Just take the first letter else: initials = ''.join([part[0] for part in name_parts]) # Take the first letter of each part # Update the column name updated_columns.append(f'Student Attendance [{initials}]') else: updated_columns.append(col) else: updated_columns.append(col) df.columns = updated_columns return df def compute_intervention_statistics(df): # Total Number of Days Available total_days = len(df) # Intervention Sessions Held sessions_held = df[INTERVENTION_COLUMN].str.strip().str.lower().eq('yes').sum() # Intervention Sessions Not Held sessions_not_held = df[INTERVENTION_COLUMN].str.strip().str.lower().eq('no').sum() # Intervention Frequency (%) intervention_frequency = (sessions_held / total_days) * 100 if total_days > 0 else 0 intervention_frequency = round(intervention_frequency, 2) # Reorder columns as specified stats = { 'Intervention Frequency (%)': [intervention_frequency], 'Intervention Sessions Held': [sessions_held], 'Intervention Sessions Not Held': [sessions_not_held], 'Total Number of Days Available': [total_days] } stats_df = pd.DataFrame(stats) return stats_df def plot_intervention_statistics(intervention_stats): # Create a stacked bar chart for sessions held and not held sessions_held = intervention_stats['Intervention Sessions Held'].values[0] sessions_not_held = intervention_stats['Intervention Sessions Not Held'].values[0] fig, ax = plt.subplots() ax.bar(['Intervention Sessions'], [sessions_held], bottom=[sessions_not_held], label='Held', color='#358E66') ax.bar(['Intervention Sessions'], [sessions_not_held], label='Not Held', color='#91D6B8') # Display values on the bars ax.text(0, sessions_not_held / 2, str(sessions_not_held), ha='center', va='center', color='white') ax.text(0, sessions_not_held + sessions_held / 2, str(sessions_held), ha='center', va='center', color='black') # Update chart settings ax.set_ylabel('Frequency') ax.legend() st.pyplot(fig) return fig def compute_student_metrics(df): # Filter DataFrame for sessions where intervention happened intervention_df = df[df[INTERVENTION_COLUMN].str.strip().str.lower() == 'yes'] intervention_sessions_held = len(intervention_df) # Get list of student columns student_columns = [col for col in df.columns if col.startswith('Student Attendance')] student_metrics = {} for col in student_columns: student_name = col.replace('Student Attendance [', '').replace(']', '').strip() # Get the attendance data for the student student_data = intervention_df[[col]].copy() # Treat blank entries as 'Absent' student_data[col] = student_data[col].fillna('Absent') # Assign attendance values attendance_values = student_data[col].apply(lambda x: 1 if x in [ ENGAGED_STR, PARTIALLY_ENGAGED_STR, NOT_ENGAGED_STR ] else 0) # Number of Sessions Attended sessions_attended = attendance_values.sum() # Attendance (%) attendance_pct = (sessions_attended / intervention_sessions_held) * 100 if intervention_sessions_held > 0 else 0 attendance_pct = round(attendance_pct, 2) # For engagement calculation, include only sessions where attendance is not 'Absent' valid_engagement_indices = attendance_values[attendance_values == 1].index engagement_data = student_data.loc[valid_engagement_indices, col] # Assign engagement values engagement_values = engagement_data.apply(lambda x: 1 if x == ENGAGED_STR else 0.5 if x == PARTIALLY_ENGAGED_STR else 0) # Sum of Engagement Values sum_engagement_values = engagement_values.sum() # Number of Sessions Attended for engagement (should be same as sessions_attended) number_sessions_attended = len(valid_engagement_indices) # Engagement (%) engagement_pct = (sum_engagement_values / number_sessions_attended) * 100 if number_sessions_attended > 0 else 0 engagement_pct = round(engagement_pct, 2) # Store metrics student_metrics[student_name] = { 'Attendance (%)': attendance_pct, 'Engagement (%)': engagement_pct } # Create a DataFrame from student_metrics student_metrics_df = pd.DataFrame.from_dict(student_metrics, orient='index').reset_index() student_metrics_df.rename(columns={'index': 'Student'}, inplace=True) return student_metrics_df def plot_student_metrics(student_metrics_df): # Create a line graph for attendance and engagement fig, ax = plt.subplots() # Plotting Attendance and Engagement with specific colors ax.plot(student_metrics_df['Student'], student_metrics_df['Attendance (%)'], marker='o', color='#005288', label='Attendance (%)') ax.plot(student_metrics_df['Student'], student_metrics_df['Engagement (%)'], marker='o', color='#3AB0FF', label='Engagement (%)') ax.set_xlabel('Student') ax.set_ylabel('Percentage (%)') ax.set_title('Student Attendance and Engagement Metrics') ax.legend() plt.xticks(rotation=45) st.pyplot(fig) return fig def download_chart(fig, filename): # Create a buffer to hold the image data buffer = io.BytesIO() # Save the figure to the buffer fig.savefig(buffer, format='png') # Set the file pointer to the beginning buffer.seek(0) # Add a download button to Streamlit st.download_button(label="Download Chart", data=buffer, file_name=filename, mime='image/png') def download_llm_output(content, filename): # Create a buffer to hold the text data buffer = io.BytesIO() buffer.write(content.encode('utf-8')) buffer.seek(0) # Add a download button to Streamlit st.download_button(label="Download LLM Output", data=buffer, file_name=filename, mime='text/plain') def prepare_llm_input(student_metrics_df): # Convert the student metrics DataFrame to a string metrics_str = student_metrics_df.to_string(index=False) llm_input = f""" Based on the following student metrics: {metrics_str} Provide: 1. Notes and Key Takeaways: Summarize the data, highlight students with the lowest and highest attendance and engagement percentages, identify students who may need adjustments to their intervention due to low attendance or engagement, and highlight students who are showing strong performance. 2. Recommendations and Next Steps: Provide interpretations based on the analysis and suggest possible next steps or strategies to improve student outcomes. """ return llm_input def prompt_response_from_hf_llm(llm_input): # Generate the refined prompt using Hugging Face API response = client.chat.completions.create( model="meta-llama/Llama-3.1-70B-Instruct", messages=[ {"role": "user", "content": llm_input} ], stream=True, temperature=0.5, max_tokens=1024, top_p=0.7 ) # Combine messages if response is streamed response_content = "" for message in response: response_content += message.choices[0].delta.content return response_content.strip() if __name__ == '__main__': main()