Tobaly's picture
Update app.py
876ad56 verified
Raw History Blame Contribute Delete
38.5 kB
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from datetime import datetime, timedelta
from sklearn.linear_model import LinearRegression
import warnings
import re
import io
import time
from difflib import SequenceMatcher
import gradio as gr
import os
import tempfile
import json
# Suppress warnings for a cleaner output
warnings.filterwarnings('ignore')
# Attempt to import Google Generative AI
try:
import google.generativeai as genai
IS_GENAI_AVAILABLE = True
except ImportError:
IS_GENAI_AVAILABLE = False
print("Warning: google-generativeai library not found. Title translation will be disabled.")
def get_now():
"""A simple function to get the current time."""
return datetime.now()
# ==============================================================================
# ComprehensivePodcastReport Class
# ==============================================================================
class ComprehensivePodcastReport:
"""
Generates a complete podcast performance report, combining overall analytics
with detailed daily and cumulative analysis of episodes within a specified period.
"""
def __init__(self, client_name):
"""Initialize the report generator."""
self.client_name = client_name
self.all_time_data = {}
self.period_data = {}
self.episode_specific_data = {}
self.unmatched_episode_csvs = []
self.title_translation_cache = {}
self.report_date = get_now().strftime("%B %d, %Y")
self.date_range_str = "All Time"
self.start_date = None
self.end_date = None
self.gemini_configured = False
def configure_gemini(self, api_key):
"""Configure Gemini API with the provided key."""
if IS_GENAI_AVAILABLE and api_key:
try:
genai.configure(api_key=api_key)
self.gemini_configured = True
print("✅ Gemini API Key configured.")
return True
except Exception as e:
print(f"❌ Error configuring Gemini API: {e}")
return False
return False
def _fix_spotify_timestamps(self, timestamp_series):
"""Fix Spotify's invalid '24:00:00' timestamps."""
return timestamp_series.astype(str).str.replace('24:00:00', '23:59:59')
def load_general_data(self, file_paths):
"""Load all general CSV files (OnSpotify, EpisodeRankings, etc.)."""
print("📊 Loading and processing general data files...")
try:
self.all_time_data['time_series'] = pd.read_csv(file_paths['time_series'])
self.all_time_data['time_series']['Date'] = pd.to_datetime(self._fix_spotify_timestamps(self.all_time_data['time_series']['Time (UTC)']))
self.all_time_data['episodes'] = pd.read_csv(file_paths['episodes'])
self.all_time_data['episodes']['Publish Date'] = pd.to_datetime(self._fix_spotify_timestamps(self.all_time_data['episodes']['Publish Time (UTC)'])).dt.normalize()
except KeyError as e:
raise Exception(f"❌ Missing a required file: {e}. Please include OnSpotify and EpisodeRankings files.")
except Exception as e:
raise Exception(f"❌ Error loading required files: {e}")
# Load optional files
optional_files = ['streams', 'geo', 'apps', 'devices']
for file_type in optional_files:
if file_type in file_paths:
try:
self.all_time_data[file_type] = pd.read_csv(file_paths[file_type])
if file_type == 'streams':
self.all_time_data['streams']['Date'] = pd.to_datetime(self._fix_spotify_timestamps(self.all_time_data['streams']['Time (UTC)']))
streams_df = self.all_time_data['streams']
first_nonzero_idx = streams_df[streams_df['Streams & Downloads'] > 0].index.min()
if pd.notna(first_nonzero_idx):
self.all_time_data['streams'] = streams_df.loc[first_nonzero_idx:].reset_index(drop=True)
print(f"✅ Loaded optional file: {file_type}")
except Exception as e:
print(f"⚠️ Warning: Could not load {file_type} file: {e}")
self.period_data = {key: df.copy() for key, df in self.all_time_data.items()}
print(f"✅ General data loaded successfully for {self.client_name}")
def filter_data_by_date(self, start_date, end_date):
"""Filter the loaded data to a specific date range."""
self.start_date = pd.to_datetime(start_date).normalize()
self.end_date = pd.to_datetime(end_date).normalize()
self.date_range_str = f"{self.start_date.strftime('%Y-%m-%d')} to {self.end_date.strftime('%Y-%m-%d')}"
print(f"🔎 Filtering data for period: {self.date_range_str}")
for key in ['time_series', 'streams']:
if key in self.period_data:
df = self.period_data[key]
self.period_data[key] = df[(df['Date'] >= self.start_date) & (df['Date'] <= self.end_date)]
def _normalize_text(self, text):
"""Normalizes text for fuzzy string matching."""
return re.sub(r'[^\u0590-\u05FFa-zA-Z0-9\-]', '', str(text).replace(':', '').replace(' ', '')).lower()
def identify_episodes_in_period(self):
"""Identifies episodes in the period and returns a formatted string and the dataframe."""
episodes_df = self.all_time_data.get('episodes')
if episodes_df is None:
return "⚠️ EpisodeRankings data is missing.", pd.DataFrame()
episodes_in_period = episodes_df[
(episodes_df['Publish Date'] >= self.start_date) &
(episodes_df['Publish Date'] <= self.end_date)
].sort_values('Publish Date', ascending=False).reset_index(drop=True)
if episodes_in_period.empty:
return ("✅ No episodes were published in the selected period. "
"You can proceed to generate the report for general stats."), episodes_in_period
output_str = f"📺 Found **{len(episodes_in_period)}** episodes published in the period. Please upload their individual CSVs below:\n\n"
for _, row in episodes_in_period.iterrows():
output_str += f"- **{row['Episode Title']}** (Published: {row['Publish Date']:%Y-%m-%d})\n"
return output_str, episodes_in_period
def load_and_match_episode_csvs(self, uploaded_files, episodes_in_period):
"""Matches and loads individual episode CSVs from uploaded files."""
if not uploaded_files or episodes_in_period.empty:
return
print("\n🔍 Matching and loading individual episode CSVs...")
episodes_in_period['normalized_title'] = episodes_in_period['Episode Title'].apply(self._normalize_text)
for file_obj in uploaded_files:
if file_obj is None:
continue
filename = os.path.basename(file_obj.name)
file_type = 'spotify' if 'OnSpotify' in filename else 'all_platforms' if 'StreamsAndDownloads' in filename else None
if not file_type:
continue
clean_filename = re.sub(r'\s*\(\d+\)|\s+\d+(?=\.csv)', '', filename)
normalized_filename = self._normalize_text(clean_filename)
episodes_in_period['similarity'] = episodes_in_period['normalized_title'].apply(
lambda title: SequenceMatcher(None, normalized_filename, title).ratio()
)
best_match = episodes_in_period.loc[episodes_in_period['similarity'].idxmax()]
if best_match['similarity'] > 0.5:
episode_title = best_match['Episode Title']
print(f" ✅ Matched: {filename[:40]}... -> {episode_title[:40]}...")
try:
# Read the file content
df = pd.read_csv(file_obj.name)
df['Date'] = pd.to_datetime(df['Time (UTC)'].str.split(',').str[0], errors='coerce').dt.normalize()
value_col = next((col for col in ['Plays', 'Streams & Downloads'] if col in df.columns), None)
if not value_col:
continue
episode_df = df[['Date', value_col]].rename(columns={value_col: 'value'}).groupby('Date')['value'].sum().reset_index()
if episode_title not in self.episode_specific_data:
self.episode_specific_data[episode_title] = {}
self.episode_specific_data[episode_title][file_type] = episode_df
except Exception as e:
print(f" -> Error processing file {filename}: {e}")
else:
self.unmatched_episode_csvs.append(filename)
def calculate_key_metrics(self, use_all_time=False):
"""Calculate core performance metrics."""
data_scope = self.all_time_data if use_all_time else self.period_data
ts_data = data_scope.get('time_series', pd.DataFrame())
if ts_data.empty:
return {}
ts_followers_cleaned = ts_data.dropna(subset=['Followers'])
current_followers = int(ts_followers_cleaned['Followers'].iloc[-1]) if not ts_followers_cleaned.empty else 0
total_plays = ts_data['Plays'].sum()
total_streams = ts_data['Streams'].sum()
metrics = {
'total_plays': int(total_plays),
'total_streams': int(total_streams),
'avg_daily_plays': ts_data['Plays'].mean(),
'engagement_percentage': (total_streams / max(total_plays, 1)) * 100,
'current_followers': current_followers
}
if use_all_time:
ep_data = self.all_time_data.get('episodes', pd.DataFrame())
metrics.update({
'total_episodes': len(ep_data),
'median_streams_per_episode': ep_data['Streams & Downloads'].median() if not ep_data.empty else 0,
'performance_benchmark': ep_data['Streams & Downloads'].quantile(0.75) if not ep_data.empty else 0
})
else:
ep_data_all = self.all_time_data.get('episodes', pd.DataFrame())
episodes_in_period = ep_data_all[
(ep_data_all['Publish Date'] >= self.start_date) &
(ep_data_all['Publish Date'] <= self.end_date)
]
metrics['performance_benchmark_period'] = episodes_in_period['Streams & Downloads'].quantile(0.75) if not episodes_in_period.empty else 0
return metrics
def analyze_growth_trends(self, period_days=30):
"""Analyze growth trends for the selected period."""
ts_data = self.period_data['time_series'].dropna(subset=['Plays', 'Followers']).sort_values('Date')
if len(ts_data) < period_days * 2:
return {}
recent_period = ts_data.tail(period_days)
previous_period = ts_data.iloc[-(period_days*2):-period_days]
if recent_period.empty or previous_period.empty:
return {}
recent_gain = recent_period['Followers'].iloc[-1] - recent_period['Followers'].iloc[0]
previous_gain = previous_period['Followers'].iloc[-1] - previous_period['Followers'].iloc[0]
return {
'plays_growth_rate_median': ((recent_period['Plays'].median() - previous_period['Plays'].median()) / max(previous_period['Plays'].median(), 1)) * 100,
'followers_gained': int(recent_gain),
'follower_growth_rate': ((recent_gain - previous_gain) / max(abs(previous_gain), 1)) * 100 if previous_gain != 0 else float('inf')
}
def analyze_seasonal_patterns(self):
"""Analyze weekly listening patterns."""
ts_data = self.period_data['time_series'].dropna(subset=['Plays'])
if len(ts_data) < 7:
return {}
ts_data['DayOfWeek'] = ts_data['Date'].dt.day_name()
weekly_avg = ts_data.groupby('DayOfWeek')['Plays'].mean()
return {
'best_day': weekly_avg.idxmax(),
'worst_day': weekly_avg.idxmin(),
'weekly_consistency': (1 - (weekly_avg.std() / weekly_avg.mean())) * 100 if weekly_avg.mean() > 0 else 0
}
def _get_platform_share(self, platform_name):
"""Helper to get platform share from all-time data."""
apps_df = self.all_time_data.get('apps')
if apps_df is None:
return 0.0
platform_row = apps_df[apps_df['App'] == platform_name]
return platform_row['Percent of Plays'].iloc[0] * 100 if not platform_row.empty else 0.0
def _add_regression_line(self, ax, x_values, y_values, label_prefix="Trend"):
"""Helper to add a linear regression trendline to a plot."""
if len(x_values) < 2:
return
x_numeric = np.arange(len(x_values))
model = LinearRegression().fit(x_numeric.reshape(-1, 1), y_values)
y_pred = model.predict(x_numeric.reshape(-1, 1))
label = f"{label_prefix} (Slope: {model.coef_[0]:.2f})"
ax.plot(x_values, y_pred, '--', color='red', alpha=0.7, label=label)
def generate_all_visualizations(self):
"""Generates a comprehensive set of all overview plots."""
print("🎨 Generating all performance visualizations...")
plots = []
ts_data = self.period_data['time_series'].sort_values('Date')
# Daily Plays
if not ts_data.empty:
fig, ax = plt.subplots(figsize=(12, 6))
ax.plot(ts_data['Date'], ts_data['Plays'], label='Daily Plays (Spotify)', color='blue')
self._add_regression_line(ax, ts_data['Date'], ts_data['Plays'])
ax.set_title(f'{self.client_name} - Daily Plays (OnSpotify)', fontsize=14)
ax.grid(True, alpha=0.3)
ax.legend()
plt.tight_layout()
plots.append(fig)
# Daily Streams
if 'streams' in self.period_data and not self.period_data['streams'].empty:
streams_data = self.period_data['streams'].sort_values('Date')
fig, ax = plt.subplots(figsize=(12, 6))
ax.plot(streams_data['Date'], streams_data['Streams & Downloads'], label='Daily Streams & Downloads', color='green')
self._add_regression_line(ax, streams_data['Date'], streams_data['Streams & Downloads'])
ax.set_title(f'{self.client_name} - Daily Streams & Downloads (All Platforms)', fontsize=14)
ax.grid(True, alpha=0.3)
ax.legend()
plt.tight_layout()
plots.append(fig)
# Monthly Plots
graph_start = self.start_date.replace(day=1) if self.start_date.day != 1 else self.start_date
for key, col, color in [('time_series', 'Plays', 'skyblue'), ('streams', 'Streams & Downloads', 'lightgreen')]:
if key in self.all_time_data:
monthly_data = self.all_time_data[key]
monthly_data = monthly_data[
(monthly_data['Date'] >= graph_start) &
(monthly_data['Date'] <= self.end_date)
]
if not monthly_data.empty:
monthly_data['YearMonth'] = monthly_data['Date'].dt.to_period('M')
monthly_values = monthly_data.groupby('YearMonth')[col].sum()
fig, ax = plt.subplots(figsize=(12, 6))
bars = ax.bar(np.arange(len(monthly_values)), monthly_values.values, color=color, edgecolor='navy')
self._add_regression_line(ax, np.arange(len(monthly_values)), monthly_values.values)
ax.set_xticks(np.arange(len(monthly_values)))
ax.set_xticklabels([str(p) for p in monthly_values.index], rotation=45)
ax.set_title(f'{self.client_name} - Monthly {col}', fontsize=14)
ax.legend()
plt.tight_layout()
current_month = pd.Period(get_now(), 'M')
if current_month in monthly_values.index:
bars[monthly_values.index.get_loc(current_month)].set_color('orange')
plots.append(fig)
# Follower Growth
follower_data = ts_data.dropna(subset=['Followers'])
if not follower_data.empty:
fig, ax = plt.subplots(figsize=(12, 6))
ax.plot(follower_data['Date'], follower_data['Followers'], label='Followers', color='purple')
self._add_regression_line(ax, follower_data['Date'], follower_data['Followers'], label_prefix="Follower Trend")
ax.set_title(f'{self.client_name} - Follower Growth', fontsize=14)
ax.grid(True, alpha=0.3)
ax.legend()
plt.tight_layout()
plots.append(fig)
# Consumption Hours
consump_data = ts_data.dropna(subset=['Consumption Hours'])
if not consump_data.empty and consump_data['Consumption Hours'].sum() > 0:
fig, ax = plt.subplots(figsize=(12, 6))
ax.plot(consump_data['Date'], consump_data['Consumption Hours'], label='Hours', color='teal')
self._add_regression_line(ax, consump_data['Date'], consump_data['Consumption Hours'], label_prefix="Consumption Trend")
ax.set_title(f'{self.client_name} - Daily Consumption Hours', fontsize=14)
ax.grid(True, alpha=0.3)
ax.legend()
plt.tight_layout()
plots.append(fig)
# Stream/Play Ratio
ratio_df = ts_data[['Date', 'Streams', 'Plays']].copy()
ratio_df['ratio'] = ratio_df['Streams'] / ratio_df['Plays'].replace(0, np.nan)
ratio_df['ratio_30d_avg'] = ratio_df['ratio'].rolling(30, min_periods=1).mean()
ratio_df.dropna(subset=['ratio_30d_avg'], inplace=True)
if not ratio_df.empty:
fig, ax = plt.subplots(figsize=(12, 6))
ax.plot(ratio_df['Date'], ratio_df['ratio_30d_avg'], label='30-Day Avg Stream/Play Ratio', color='orange')
self._add_regression_line(ax, ratio_df['Date'], ratio_df['ratio_30d_avg'], label_prefix="Ratio Trend")
ax.set_title(f'{self.client_name} - Stream/Play Ratio (30-Day Average)', fontsize=14)
ax.grid(True, alpha=0.3)
ax.legend()
plt.tight_layout()
plots.append(fig)
# Platform Distribution
if 'apps' in self.all_time_data:
fig, ax = plt.subplots(figsize=(8, 8))
self.all_time_data['apps'].head(5).plot.pie(
y='Percent of Plays',
labels=self.all_time_data['apps']['App'].head(5),
autopct='%1.1f%%',
ax=ax,
legend=False
)
ax.set_title(f'{self.client_name} - Platform Distribution', fontsize=14)
ax.set_ylabel('')
plt.tight_layout()
plots.append(fig)
# Device Distribution
if 'devices' in self.all_time_data:
fig, ax = plt.subplots(figsize=(8, 8))
self.all_time_data['devices'].head(5).plot.pie(
y='Percent of Plays',
labels=self.all_time_data['devices']['Device'].head(5),
autopct='%1.1f%%',
ax=ax,
legend=False
)
ax.set_title(f'{self.client_name} - Device Distribution', fontsize=14)
ax.set_ylabel('')
plt.tight_layout()
plots.append(fig)
# Geographic Markets
if 'geo' in self.all_time_data:
fig, ax = plt.subplots(figsize=(10, 6))
top_geo = self.all_time_data['geo'].head(10)
ax.barh(top_geo['Geo'], top_geo['Percent of Plays'] * 100, color='coral')
ax.set_title(f'{self.client_name} - Top 10 Geographic Markets', fontsize=14)
ax.invert_yaxis()
ax.grid(True, alpha=0.3, axis='x')
plt.tight_layout()
plots.append(fig)
# Episode Performance Distribution
ep_data_all = self.all_time_data.get('episodes')
if ep_data_all is not None:
fig, ax = plt.subplots(figsize=(10, 6))
ax.hist(ep_data_all['Streams & Downloads'], bins=30, color='skyblue', edgecolor='navy')
ax.axvline(ep_data_all['Streams & Downloads'].mean(), color='red', linestyle='--', label='Mean')
ax.set_title(f'{self.client_name} - Episode Performance Distribution (All Time)', fontsize=14)
ax.legend()
plt.tight_layout()
plots.append(fig)
episodes_in_period = ep_data_all[
(ep_data_all['Publish Date'] >= self.start_date) &
(ep_data_all['Publish Date'] <= self.end_date)
]
if not episodes_in_period.empty:
fig, ax = plt.subplots(figsize=(10, 6))
ax.hist(episodes_in_period['Streams & Downloads'], bins=15, color='teal', edgecolor='darkslategray')
ax.axvline(episodes_in_period['Streams & Downloads'].mean(), color='red', linestyle='--', label='Mean')
ax.set_title(f'{self.client_name} - Episode Performance (Published in Period)', fontsize=14)
ax.legend()
plt.tight_layout()
plots.append(fig)
return plots
def _translate_title(self, hebrew_title):
"""Translates a Hebrew title to a short English legend title using Gemini."""
if hebrew_title in self.title_translation_cache:
return self.title_translation_cache[hebrew_title]
if not self.gemini_configured:
return hebrew_title[:40] + '...'
try:
model = genai.GenerativeModel('gemini-2.5-flash')
prompt = f"""Analyze the Hebrew podcast title "{hebrew_title}". Generate a short, readable English legend title formatted as "[Guest Name] - #[Episode Number]". Transliterate the name. If there is no episode number, use only the name. If there is no name, create a concise English summary of the topic. If you recognize מהדורת מוסף in the title, then the output would be: "Investors Panel - #[Episode Number]".The final title must be under 5 words."""
response = model.generate_content(prompt)
time.sleep(1.5)
translated = response.text.strip()
self.title_translation_cache[hebrew_title] = translated
return translated
except Exception as e:
print(f" -> Gemini API Error: {e}")
return hebrew_title[:40] + '...'
def generate_episode_specific_plots(self, max_episodes=10):
"""Generates the cumulative performance plots for individual episodes."""
print("🚀 Generating episode-specific cumulative plots...")
plots = []
if not self.episode_specific_data:
return plots
for data_source in ['spotify', 'all_platforms']:
ep_data_source = {t: d[data_source] for t, d in self.episode_specific_data.items() if data_source in d}
if not ep_data_source:
continue
ep_totals = {title: df['value'].sum() for title, df in ep_data_source.items()}
top_episodes = sorted(ep_totals, key=ep_totals.get, reverse=True)[:max_episodes]
date_index = pd.date_range(start=self.start_date, end=self.end_date, freq='D')
plot_df = pd.DataFrame(index=date_index)
for title in top_episodes:
plot_df = plot_df.join(ep_data_source[title].set_index('Date').rename(columns={'value': title}))
cumulative_df = plot_df.fillna(0).cumsum()
fig, ax = plt.subplots(figsize=(15, 8))
for h_title in cumulative_df.columns:
ax.plot(cumulative_df.index, cumulative_df[h_title], label=self._translate_title(h_title), linewidth=2.5)
ax.set_title(f'Cumulative Performance by Date ({data_source.replace("_", " ").title()})', fontsize=16)
ax.legend(bbox_to_anchor=(1.02, 1), loc='upper left')
plt.tight_layout()
plots.append(fig)
# Air day analysis
airday_df = pd.DataFrame()
for title in top_episodes:
df = ep_data_source[title]
first_play = df[df['value'] > 0]['Date'].min()
if pd.isna(first_play):
continue
df_after = df[df['Date'] >= first_play].copy()
airday_df[title] = df_after['value'].cumsum().reset_index(drop=True)
if not airday_df.empty:
airday_df['Average'] = airday_df.mean(axis=1)
fig, ax = plt.subplots(figsize=(15, 8))
for h_title in top_episodes:
if h_title in airday_df.columns:
ax.plot(airday_df.index + 1, airday_df[h_title], label=self._translate_title(h_title), alpha=0.8)
ax.plot(airday_df.index + 1, airday_df['Average'], 'k--', label='Average', linewidth=2.5)
ax.set_title(f'Cumulative Performance by Air Day ({data_source.replace("_", " ").title()})', fontsize=16)
ax.set_xlabel('Days Since Release')
ax.legend(bbox_to_anchor=(1.02, 1), loc='upper left')
plt.tight_layout()
plots.append(fig)
return plots
def generate_executive_summary(self, all_time_metrics, period_metrics, trends, seasonal):
"""Generates the final executive summary text."""
summary = f"""# {self.client_name} - Podcast Performance Report
*Generated on {self.report_date}*
---
## 📈 All-Time Performance Summary
*A high-level overview of the podcast's entire history.*
- **Total Plays**: {all_time_metrics.get('total_plays', 0):,}
- **Total Streams**: {all_time_metrics.get('total_streams', 0):,}
- **Total Episodes**: {all_time_metrics.get('total_episodes', 0)}
- **Most Recent Follower Count**: {all_time_metrics.get('current_followers', 0):,}
- **Median Streams per Episode**: {all_time_metrics.get('median_streams_per_episode', 0):,.0f}
- **Performance Benchmark (Top 25% of Episodes)**: > {all_time_metrics.get('performance_benchmark', 0):,.0f} streams
---
## 📊 Performance for Selected Period: {self.date_range_str}
*A detailed analysis of performance within the specified date range.*
### Key Performance Indicators (Period)
- **Total Plays in Period**: {period_metrics.get('total_plays', 0):,}
- **Total Streams in Period**: {period_metrics.get('total_streams', 0):,}
- **Average Daily Plays**: {period_metrics.get('avg_daily_plays', 0):,.0f}
- **Engagement Rate (Streams/Plays)**: {period_metrics.get('engagement_percentage', 0):.1f}%
- **Performance Benchmark (Top 25% of Episodes in Period)**: > {period_metrics.get('performance_benchmark_period', 0):,.0f} streams
"""
if trends:
summary += f"""
### Growth Trends (Comparison of 30-Day Sub-periods)
- **Plays Growth (Median)**: {trends.get('plays_growth_rate_median', 0):+.1f}%
- **Followers Gained**: {trends.get('followers_gained', 0):+,}
- **Follower Growth Rate**: {trends.get('follower_growth_rate', 0):+.1f}% (vs. prior period)
"""
if seasonal:
summary += f"""
### Listening Patterns (Period)
- **Best Day for Plays**: {seasonal.get('best_day')}
- **Worst Day for Plays**: {seasonal.get('worst_day')}
- **Weekly Listening Consistency**: {seasonal.get('weekly_consistency', 0):.1f}%
"""
summary += "\n### Audience Insights (All-Time)\n"
if 'geo' in self.all_time_data:
summary += f"- **Primary Market**: {self.all_time_data['geo'].loc[0, 'Geo']} ({self.all_time_data['geo'].loc[0, 'Percent of Plays'] * 100:.1f}% of plays)\n"
if 'apps' in self.all_time_data:
summary += f"- **Platform Mix**: {self._get_platform_share('Spotify'):.1f}% Spotify, {self._get_platform_share('Apple Podcasts'):.1f}% Apple Podcasts\n"
summary += "\n## 🏆 Top 5 Performing Episodes (All-Time)\n"
ep_data = self.all_time_data.get('episodes')
if ep_data is not None:
for i, (_, row) in enumerate(ep_data.nlargest(5, 'Streams & Downloads').iterrows(), 1):
summary += f"{i}. **{row['Episode Title'][:60]}...**\n {row['Streams & Downloads']:,} streams | {row['Publish Date']:%B %Y}\n\n"
return summary
# ==============================================================================
# Gradio specific functions
# ==============================================================================
def process_step_1(api_key, client_name, date_from_str, date_to_str, general_files):
"""Handles the first step of the analysis."""
try:
if not client_name or not date_from_str or not general_files:
return None, "❌ Please fill in all required fields and upload files.", None, gr.update(visible=False)
# Initialize analyzer
analyzer = ComprehensivePodcastReport(client_name)
# Configure Gemini if API key provided
if api_key:
analyzer.configure_gemini(api_key)
# Process uploaded files
file_patterns = {
'time_series': 'OnSpotify',
'episodes': 'EpisodeRankings',
'geo': 'GeoLocation',
'apps': 'PlaysByApp',
'devices': 'PlaysByDevice',
'streams': 'StreamsAndDownloads'
}
detected_files = {}
for f in general_files:
if f is None:
continue
fname = os.path.basename(f.name)
for ftype, pattern in file_patterns.items():
if pattern in fname:
detected_files[ftype] = f.name
break
if 'time_series' not in detected_files or 'episodes' not in detected_files:
return None, "❌ Required files not found. Please upload at least OnSpotify and EpisodeRankings CSVs.", None, gr.update(visible=False)
# Load data
analyzer.load_general_data(detected_files)
# Set date range
end_date = pd.to_datetime(date_to_str) if date_to_str else get_now()
try:
date_from = int(date_from_str)
start_date = end_date - timedelta(days=date_from)
except ValueError:
start_date = pd.to_datetime(date_from_str)
analyzer.filter_data_by_date(start_date, end_date)
# Identify episodes in period
episode_list_md, episodes_in_period_df = analyzer.identify_episodes_in_period()
# Convert episodes dataframe to JSON for state management
episodes_json = episodes_in_period_df.to_json(orient='records', date_format='iso')
return analyzer, episode_list_md, episodes_json, gr.update(visible=True)
except Exception as e:
error_msg = f"❌ Error during initial processing: {str(e)}"
print(error_msg)
return None, error_msg, None, gr.update(visible=False)
def process_step_2(analyzer, episodes_json, episode_files):
"""Handles the final report generation, returning a summary and a list of plots."""
try:
if analyzer is None:
return "❌ Error: Initial processing step was not completed.", []
# Reconstruct episodes dataframe from JSON
episodes_data = json.loads(episodes_json) if episodes_json else []
episodes_in_period_df = pd.DataFrame(episodes_data)
if not episodes_in_period_df.empty and 'Publish Date' in episodes_in_period_df.columns:
episodes_in_period_df['Publish Date'] = pd.to_datetime(episodes_in_period_df['Publish Date'])
# Load episode-specific data
analyzer.load_and_match_episode_csvs(episode_files or [], episodes_in_period_df)
# Calculate metrics
all_time_metrics = analyzer.calculate_key_metrics(use_all_time=True)
period_metrics = analyzer.calculate_key_metrics(use_all_time=False)
trends = analyzer.analyze_growth_trends()
seasonal = analyzer.analyze_seasonal_patterns()
# Generate summary
summary = analyzer.generate_executive_summary(all_time_metrics, period_metrics, trends, seasonal)
# Generate plots
general_plots = analyzer.generate_all_visualizations()
episode_plots = analyzer.generate_episode_specific_plots()
all_plots = general_plots + episode_plots
# Convert matplotlib figures to images for Gradio
plot_images = []
for fig in all_plots:
temp_file = tempfile.NamedTemporaryFile(delete=False, suffix='.png')
fig.savefig(temp_file.name, dpi=150, bbox_inches='tight')
plt.close(fig) # Important: close figure to free memory
plot_images.append(temp_file.name)
return summary, plot_images
except Exception as e:
error_msg = f"❌ Error during report generation: {str(e)}"
print(error_msg)
return error_msg, []
# ==============================================================================
# Gradio UI Layout
# ==============================================================================
with gr.Blocks(theme=gr.themes.Soft(), title="Valuebell Podcast Analytics") as demo:
gr.HTML("""
<div style="text-align: center; margin-bottom: 30px;">
<h1>🎙️ Valuebell Podcast Analytics Dashboard</h1>
<p style="font-size: 18px; color: #666;">
Professional podcast performance reports from Spotify for Creators CSV exports
</p>
</div>
""")
# State variables
analyzer_state = gr.State(None)
episodes_df_state = gr.State(None)
with gr.Row():
with gr.Column(scale=1):
gr.Markdown("## 📋 Step 1: Initial Setup")
with gr.Group():
api_key_input = gr.Textbox(
label="🔑 Gemini API Key (Optional)",
type="password",
placeholder="For Hebrew title translation...",
info="Leave empty if not using Hebrew content"
)
client_name_input = gr.Textbox(
label="🎙️ Client / Podcast Name",
placeholder="e.g., The Co-Pilot Podcast",
value="My Podcast"
)
with gr.Row():
date_from_input = gr.Textbox(
label="📅 Start Date or Period",
placeholder="e.g., 60 (last 60 days) or 2025-07-01",
value="60"
)
date_to_input = gr.Textbox(
label="📅 End Date (Optional)",
placeholder="e.g., 2025-08-18 (defaults to today)"
)
general_files_input = gr.File(
label="📁 Upload General CSVs",
file_count="multiple",
file_types=[".csv"]
)
gr.HTML("""
<div style="background-color: #f0f8ff; padding: 10px; border-radius: 5px; margin: 10px 0;">
<small><strong>Required:</strong> OnSpotify, EpisodeRankings<br>
<strong>Optional:</strong> StreamsAndDownloads, GeoLocation, PlaysByApp, PlaysByDevice</small>
</div>
""")
step1_button = gr.Button("🚀 1. Process General Files", variant="primary", size="lg")
with gr.Column(scale=2):
with gr.Group(visible=False) as step2_box:
gr.Markdown("## 📊 Step 2: Episode Details & Report Generation")
episode_list_output = gr.Markdown("Processing...")
episode_files_input = gr.File(
label="📁 Upload Individual Episode CSVs (Optional)",
file_count="multiple",
file_types=[".csv"]
)
step2_button = gr.Button("📈 2. Generate Full Report", variant="primary", size="lg")
gr.Markdown("---")
with gr.Column():
gr.Markdown("## 📋 Executive Summary")
summary_output = gr.Markdown("Upload files and process to see your comprehensive report...")
gr.Markdown("## 📊 Performance Visualizations")
gallery_output = gr.Gallery(
label="Generated Charts",
columns=2,
object_fit="contain",
height="auto",
show_label=False
)
# Event handlers
step1_button.click(
fn=process_step_1,
inputs=[api_key_input, client_name_input, date_from_input, date_to_input, general_files_input],
outputs=[analyzer_state, episode_list_output, episodes_df_state, step2_box],
show_progress=True
)
step2_button.click(
fn=process_step_2,
inputs=[analyzer_state, episodes_df_state, episode_files_input],
outputs=[summary_output, gallery_output],
show_progress=True
)
gr.HTML("""
<div style="text-align: center; margin-top: 30px; padding: 20px; background-color: #f8f9fa; border-radius: 8px;">
<h4>💡 Tips for Best Results:</h4>
<ul style="text-align: left; display: inline-block; max-width: 600px;">
<li>Ensure CSV files contain the expected keywords in their names</li>
<li>For Hebrew podcasts, add your Gemini API key for title translation</li>
<li>Include episode-specific CSV files for detailed performance tracking</li>
<li>Use date ranges that cover complete weeks for better trend analysis</li>
</ul>
</div>
""")
if __name__ == "__main__":
demo.launch(share=True, debug=True)