app.py

import streamlit as st
import pandas as pd
import random
import time
import string
import gspread
import os
import json
import datetime
from oauth2client.service_account import ServiceAccountCredentials

# Load worker-specific stimuli
def get_google_creds():
    service_account_json = os.getenv("SERVICE_ACCOUNT_JSON")
    if service_account_json:
        creds_dict = json.loads(service_account_json)
        scope = ["https://spreadsheets.google.com/feeds", "https://www.googleapis.com/auth/drive"]
        creds = ServiceAccountCredentials.from_json_keyfile_dict(creds_dict, scope)
        return gspread.authorize(creds)
    else:
        st.error("Google service account credentials not found.")
        return None

def get_next_worker_id():
    client = get_google_creds()
    if client is None:
        return None
    
    try:
        # sheet_name = "Odd-One-Out Experiment Responses"
        sheet_name = "Odd-one-out results v2.0"
        sheet = client.open(sheet_name).sheet1
        existing_worker_ids = sheet.col_values(1)[1:]  # Get worker IDs (skip header)
        assigned_workers = set(map(int, existing_worker_ids))
        
        for i in range(1, 21):  # Assign worker ID between 1 and 20
            if i not in assigned_workers:
                return i
        return None  # No available worker slots
    except Exception as e:
        st.error(f"Error retrieving worker ID: {str(e)}")
        return None

# Load worker-specific stimuli
@st.cache_data
def load_worker_data(worker_id):
    file_path = os.path.join("exp1", f"worker_{worker_id:02d}.tsv")
    if os.path.exists(file_path):
        df = pd.read_csv(file_path, sep='\t')
        return df.dropna().reset_index(drop=True)
    else:
        return None

# Automatically assign worker ID
if 'worker_id' not in st.session_state:
    # assigned_worker_id = get_next_worker_id()
    assigned_worker_id = 4
    if assigned_worker_id is None:
        st.error("No available worker slots. Please try again later.")
        st.stop()
    st.session_state.worker_id = assigned_worker_id
    st.session_state.df = load_worker_data(st.session_state.worker_id)
    st.session_state.step = "instructions"
    st.rerun()

# Load main experiment data assigned to individual worker
df = st.session_state.df
if df is None:
    st.error("No stimuli available for this worker.")
    st.stop()

# Function to generate a unique passcode
def generate_passcode():
    return ''.join(random.choices(string.ascii_uppercase + string.digits, k=8))

# Function to upload responses to Google Drive
def upload_to_google_drive(response_df):
    try:
        client = get_google_creds()
        # sheet_name = "Odd-One-Out Experiment Responses"
        sheet_name = "Odd-one-out results v2.0"
        try:
            sheet = client.open(sheet_name).sheet1
        except gspread.exceptions.SpreadsheetNotFound:
            sheet = client.create(sheet_name).sheet1
            sheet.append_row(["worker_id", "passcode", "question", "keyword", "selected", "correct_answer", "is_correct", "response_time"])

        data_list = response_df.values.tolist()
        for row in data_list:
            sheet.append_row(row)

        st.success("✅ Your responses have been recorded successfully.")
    except Exception as e:
        st.error(f"Error uploading to Google Drive: {str(e)}")

# Load training samples
def load_training_samples():
    file_path = "training_samples.csv"
    if os.path.exists(file_path):
        return pd.read_csv(file_path).dropna().reset_index(drop=True)
    else:
        st.error("Training samples file not found.")
        return pd.DataFrame()


# Load training data
training_df = load_training_samples()

# Initialize session state variables
if "step" not in st.session_state:
    st.session_state.step = "instructions"
if "training_index" not in st.session_state:
    st.session_state.training_index = 0
if "training_complete" not in st.session_state:
    st.session_state.training_complete = False
if "experiment_index" not in st.session_state:
    st.session_state.experiment_index = 0
if "experiment_complete" not in st.session_state:
    st.session_state.experiment_complete = False
if "responses" not in st.session_state:
    st.session_state.responses = []
if "start_time" not in st.session_state:
    st.session_state.start_time = None
if "passcode" not in st.session_state:
    st.session_state.passcode = None
if "show_answer" not in st.session_state:
    st.session_state.show_answer = False

# Increase font size for instructions and multiple-choice options
st.markdown("""
    <style>
    .stApp {
        font-size: 17px !important;
    }
    .correct-answer {
        color: green;
    }
    .highlight-red {
        color: #D9534F;
        font-weight: bold;
    }
    /* Increase font size for radio button labels */
    div[data-testid="stRadio"] label {
        font-size: 17px !important;
    }
    /* Increase font size for general text */
    div[data-testid="stMarkdownContainer"] {
        font-size: 17px !important;
    }
    div[data-testid="stVerticalBlock"] p {
        font-size: 17px !important;
    }
    </style>
    """, unsafe_allow_html=True)

st.title("Scene Identification Experiment")

# **STEP 1: Instructions Page**
if st.session_state.step == "instructions":
    st.header("📖 Instructions")
    st.write("""
        Welcome to the experiment! Here’s how it works:

        - You will be presented with five short text passages excerpted from **fictional stories**.
        - The passages describe scenes related to a **specific keyword**.
        - Four of these passages describe **a similar type of scene**, while **one is different**.
        - Your task is to identify the passage that describes <span style='color:#D9534F; font-weight:bold;'>a scene distinct from the rest</span>.
        - Please try to answer as quickly and accurately as possible.
    """, unsafe_allow_html=True)

    if st.button("Start Practicing"):
        st.session_state.step = "training"
        st.rerun()

# **STEP 2: Training Phase**
elif st.session_state.step == "training":
    if st.session_state.training_index >= len(training_df):
        st.session_state.step = "training_complete"
        st.rerun()
    else:
        row = training_df.iloc[st.session_state.training_index]
        
        st.write(f"### Practice Sample {st.session_state.training_index + 1}/{len(training_df)}")
        st.write(f"**Keyword: {row['keyword']}**")
        
        options = [row['text_1'], row['text_2'], row['text_3'], row['text_4'], row['text_5']]
        correct_answer = options[row['odd_index'] - 1]  # Get the correct answer using odd_index
        
        user_choice = st.radio("Which scene is distinct from the others?", options, 
            key=f"training_{st.session_state.training_index}", index=None)

        if st.button("Submit Answer"):
            if user_choice is None:
                st.warning("⚠️ Please select an option first!")
            else:
                st.session_state.show_answer = True
                st.rerun()

        if st.session_state.show_answer:
            st.write("### ✅ The correct answer was:")
            st.markdown(f"<p class='correct-answer'>{correct_answer}</p>", unsafe_allow_html=True)
            if st.button("Next Practice Sample"):
                st.session_state.show_answer = False
                st.session_state.training_index += 1
                st.rerun()

# **STEP 3: Training Complete Page**
elif st.session_state.step == "training_complete":
    st.header("🎉 Practice Complete!")
    # st.write("""
    # You have completed the practice phase! Now, you will proceed to the main experiment.
    # Please remember to select the passage that describes <span style='color:#D9534F; font-weight:bold;'>a distinct scene from the rest</span>.
    
    # Click **"Start Experiment"** when you are ready. Your response time will begin after you proceed.
    # """)
    st.markdown("""
        You have completed the practice phase! Now, you will proceed to the main experiment.  
        Please remember to select the passage that describes <span style='color:#D9534F; font-weight:bold;'>a distinct scene from the rest</span>.

    Click **"Start Experiment"** when you are ready. Your response time will begin after you proceed.
    """, unsafe_allow_html=True)
    
    if st.button("Start Experiment"):
        st.session_state.start_time = time.time()
        st.session_state.step = "experiment"
        st.rerun()

# Ensure 'Submit Answer' button is hidden after experiment completion
if 'experiment_complete' not in st.session_state:
    st.session_state.experiment_complete = False

# **STEP 4: Main Experiment**
elif st.session_state.step == "experiment":
    if st.session_state.experiment_index >= len(df):
        st.success("✅ You've completed the experiment!")
        st.session_state.experiment_complete = True

        if "passcode" not in st.session_state or st.session_state.passcode is None:
            st.session_state.passcode = generate_passcode()

        st.write("### 🎉 Your unique completion passcode:")
        st.code(st.session_state.passcode, language="text")

        response_df = pd.DataFrame(st.session_state.responses)
        response_df["worker_id"] = st.session_state.worker_id
        response_df["passcode"] = st.session_state.passcode

        upload_to_google_drive(response_df)
    else:
        row = df.iloc[st.session_state.experiment_index]
        
        st.write(f"### Question {st.session_state.experiment_index + 1}/{len(df)}")
        st.write(f"**Keyword: {row['keyword']}**")
        
        user_choice = st.radio("Which scene is distinct from the others?", [
            row['text_1'], row['text_2'], row['text_3'], row['text_4'], row['text_5']
        ], key=st.session_state.experiment_index, index=None)

        options = [row['text_1'], row['text_2'], row['text_3'], row['text_4'], row['text_5']]
        correct_answer = options[row['odd_index'] - 1]  # Get the correct answer using odd_index

        # if not st.session_state.experiment_complete:
        if not st.session_state.experiment_complete and st.session_state.experiment_index < len(df):
            if st.button("Submit Answer"):
                if user_choice is None:
                    st.warning("⚠️ Please select an option first!")
                else:
                    response_time = time.time() - st.session_state.start_time

                    st.session_state.responses.append({
                        'timestamp': datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
                        'worker_id': st.session_state.worker_id,
                        'passcode': st.session_state.passcode if st.session_state.passcode else "TEMP",
                        'question': st.session_state.experiment_index + 1,
                        'keyword': row['keyword'],
                        'selected': user_choice,
                        'correct_answer': correct_answer,
                        'is_correct': user_choice == correct_answer,
                        'response_time': round(response_time, 2)
                    })

                    st.session_state.experiment_index += 1
                    st.session_state.start_time = time.time()
                    st.rerun()