pages/Summary.py

import json
import os

import datasets
import numpy as np
import pandas as pd
import pymysql.cursors
import streamlit as st

from datetime import datetime
from streamlit_elements import elements, mui, html, dashboard, nivo
from streamlit_extras.switch_page_button import switch_page
from streamlit_extras.metric_cards import style_metric_cards
from streamlit_extras.stylable_container import stylable_container

from pages.Gallery import load_hf_dataset
from pages.Ranking import connect_to_db


class DashboardApp:
    def __init__(self, roster, promptBook, session_finished):
        self.roster = roster
        self.promptBook = promptBook
        self.session_finished = session_finished

    def sidebar(self, tags, mode):
        with st.sidebar:
            tag = st.selectbox('Select a tag', tags, key='tag')
            # st.write('---')
            with st.form('summary_sidebar_form'):
                st.write('## Want a more comprehensive summary?')
                st.write('Jump back to gallery and select more images to rank!')
                back_to_gallery = st.form_submit_button('🖼️ Go to Gallery')
                if back_to_gallery:
                    switch_page('gallery')
                back_to_ranking = st.form_submit_button('🎖️ Go to Ranking')
                if back_to_ranking:
                    switch_page('ranking')

            # with st.form('overall_feedback'):
            #     feedback = st.text_area('Please leave your comments here.', key='comment')
            #     submit_feedback = st.form_submit_button('Submit Feedback')
            #     if submit_feedback:
            #         print(feedback)

        return tag

    def leaderboard(self, tag, db_table):
        tag = '%' if tag == 'all' else tag

        # get the ranking results of the current user
        curser = RANKING_CONN.cursor()
        curser.execute(f"SELECT * FROM {db_table} WHERE username = '{st.session_state.user_id[0]}' AND timestamp = '{st.session_state.user_id[1]}' AND tag LIKE '{tag}'")
        results = curser.fetchall()
        curser.close()

        modelVersion_standings = self.score_calculator(results, db_table)

        # sort the modelVersion_standings by value into a list of tuples in descending order
        modelVersion_standings = sorted(modelVersion_standings.items(), key=lambda x: x[1], reverse=True)

        tab1, tab2 = st.tabs(['Top Picks', 'Detailed Info'])

        with tab1:
            # self.podium(modelVersion_standings)
            self.podium_expander(modelVersion_standings)

        with tab2:
            st.write('## Detailed information of all selected models')
            detailed_info = pd.merge(pd.DataFrame(modelVersion_standings, columns=['modelVersion_id', 'ranking_score']), self.roster, on='modelVersion_id')
            st.data_editor(detailed_info, hide_index=True, disabled=True)

    def podium(self, modelVersion_standings, n=3):
        st.write('## Top picks')
        metric_cols = st.columns(n)
        image_display = st.empty()

        for i in range(n):
            with metric_cols[i]:
                modelVersion_id = modelVersion_standings[i][0]
                winning_times = modelVersion_standings[i][1]

                model_id, model_name, modelVersion_name, url = self.roster[self.roster['modelVersion_id'] == modelVersion_id][['model_id', 'model_name', 'modelVersion_name', 'modelVersion_url']].values[0]

                metric_card = stylable_container(
                    key="container_with_border",
                    css_styles="""
                    {
                        border: 1.5px solid rgba(49, 51, 63, 0.2);
                        border-left: 0.5rem solid gold;
                        border-radius: 5px;
                        padding: calc(1em + 5px);
                        gap: 0.5em;
                        box-shadow: 0 0 2rem rgba(0, 0, 0, 0.08);
                        overflow-x: scroll;
                    }
                    """,
                )

                with metric_card:
                    icon = '🥇'if i == 0 else '🥈' if i == 1 else '🥉'
                    # st.write(model_id)
                    st.write(f'### {icon} {model_name}, [{modelVersion_name}](https://civitai.com/models/{model_id}?modelVersionId={modelVersion_id})')
                    st.write(f'Ranking Score: {winning_times}')

                show_image = st.button('Show Image', key=modelVersion_id, use_container_width=True)
                if show_image:

                    images = self.promptBook[self.promptBook['modelVersion_id'] == modelVersion_id]['image_id'].values
                    with image_display.container():
                        st.write('---')
                        st.write(f'### Images generated with {icon} {model_name}, {modelVersion_name}')
                        col_num = 4
                        image_cols = st.columns(col_num)
                        for i in range(len(images)):
                            with image_cols[i % col_num]:
                                image = f"https://modelcofferbucket.s3-accelerate.amazonaws.com/{images[i]}.png"
                                st.image(image, use_column_width=True)

    def podium_expander(self, modelVersion_standings, n=3):
        # st.write('## Top picks')
        # metric_cols = st.columns(n)
        for i in range(n):
            # with metric_cols[i]:
            modelVersion_id = modelVersion_standings[i][0]
            winning_times = modelVersion_standings[i][1]

            model_id, model_name, modelVersion_name, url = self.roster[self.roster['modelVersion_id'] == modelVersion_id][['model_id', 'model_name', 'modelVersion_name', 'modelVersion_url']].values[0]

            icon = '🥇'if i == 0 else '🥈' if i == 1 else '🥉'
            podium_display = st.columns([1, 14])
            with podium_display[0]:
                st.title(f'{icon}')
            with podium_display[1]:
                st.write(f'##### {model_name}, {modelVersion_name}')
                st.write(f'[Civitai Page](https://civitai.com/models/{model_id}?modelVersionId={modelVersion_id}), [Model Download Link]({url}), Ranking Score: {winning_times}')
                # with st.expander(f'**{icon} {model_name}, [{modelVersion_name}](https://civitai.com/models/{model_id}?modelVersionId={modelVersion_id})**, Ranking Score: {winning_times}'):
                with st.expander(f'Show Images'):
                    images = self.promptBook[self.promptBook['modelVersion_id'] == modelVersion_id]['image_id'].values

                    safety_check = st.toggle('Include potentially unsafe or offensive images', value=False, key=modelVersion_id)
                    unsafe_prompts = json.load(open('data/unsafe_prompts.json', 'r'))
                    # merge dict values into one list
                    unsafe_prompts = [item for sublist in unsafe_prompts.values() for item in sublist]
                    unsafe_images = self.promptBook[self.promptBook['prompt_id'].isin(unsafe_prompts)]['image_id'].values

                    if not safety_check:
                        # exclude unsafe prompts from images
                        images = [image for image in images if image not in unsafe_images]

                    # st.write(f'### Images generated with {icon} {model_name}, {modelVersion_name}')
                    col_num = 4
                    image_cols = st.columns(col_num)

                    for j in range(len(images)):
                        with image_cols[j % col_num]:
                            image = f"https://modelcofferbucket.s3-accelerate.amazonaws.com/{images[j]}.png"
                            st.image(image, use_column_width=True)
            if i != n - 1:
                st.write('---')

    def score_calculator(self, results, db_table):
        modelVersion_standings = {}
        if db_table == 'battle_results':
            # sort results by battle time
            results = sorted(results, key=lambda x: x['battletime'])

            for record in results:
                modelVersion_standings[record['winner']] = modelVersion_standings.get(record['winner'], 0) + 1

                # add the loser who never wins
                if record['loser'] not in modelVersion_standings:
                    modelVersion_standings[record['loser']] = 0

                # add the winning time of the loser to the winner
                modelVersion_standings[record['winner']] += modelVersion_standings[record['loser']]

        elif db_table == 'sort_results':
            pts_map = {'position1': 5, 'position2': 3, 'position3': 1, 'position4': 0}
            for record in results:
                for i in range(1, 5):
                    modelVersion_standings[record[f'position{i}']] = modelVersion_standings.get(record[f'position{i}'], 0) + pts_map[f'position{i}']

        return modelVersion_standings

    def app(self):
        st.write('### Your Preferred Models')

        # mode = st.sidebar.radio('Ranking mode', ['Drag and Sort', 'Battle'], horizontal=True, index=1)
        mode = st.session_state.assigned_rank_mode
        # get tags from database of the current user
        db_table = 'sort_results' if mode == 'Drag and Sort' else 'battle_results'

        tags = ['all']
        curser = RANKING_CONN.cursor()
        curser.execute(
            f"SELECT DISTINCT tag FROM {db_table} WHERE username = '{st.session_state.user_id[0]}' AND timestamp = '{st.session_state.user_id[1]}'")
        for row in curser.fetchall():
            tags.append(row['tag'])
        curser.close()

        if tags == ['all']:
            st.info(f'No rankings are finished with {mode} mode yet.')

        else:
            tag = self.sidebar(tags, mode)
            self.leaderboard(tag, db_table)

        comment = st.chat_input('Please leave your comments here.', key='comment')
        if comment:
            commenttime = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
            curser = RANKING_CONN.cursor()
            # parse the comment to at most 300 to avoid SQL injection
            for i in range(0, len(comment), 300):
                curser.execute(f"INSERT INTO comments (username, timestamp, comment, commenttime) VALUES ('{st.session_state.user_id[0]}', '{st.session_state.user_id[1]}', '{comment[i:i+300]}', '{commenttime}')")
            RANKING_CONN.commit()
            curser.close()

            st.toast('Thanks for your feedback! We will take it into consideration in our future work.')


if __name__ == "__main__":
    st.set_page_config(layout="wide")

    if 'user_id' not in st.session_state:
        st.warning('Please log in first.')
        home_btn = st.button('Go to Home Page')
        if home_btn:
            switch_page("home")

    elif 'progress' not in st.session_state:
        st.info('You have not checked any image yet. Please go back to the gallery page and check some images.')
        gallery_btn = st.button('🖼️ Go to Gallery')
        if gallery_btn:
            switch_page('gallery')

    else:
        session_finished = []

        for key, value in st.session_state.progress.items():
            if value == 'finished':
                session_finished.append(key)

        if len(session_finished) == 0:
            st.info('A dashboard showing your preferred models will appear after you finish any ranking session.')
            ranking_btn = st.button('🎖️ Go to Ranking')
            if ranking_btn:
                switch_page('ranking')
            gallery_btn = st.button('🖼️ Go to Gallery')
            if gallery_btn:
                switch_page('gallery')

        else:
            roster, promptBook, images_ds = load_hf_dataset(st.session_state.show_NSFW)
            RANKING_CONN = connect_to_db()
            app = DashboardApp(roster, promptBook, session_finished)
            app.app()