import streamlit as st import pandas as pd import plotly.express as px import plotly.graph_objects as go from datetime import datetime, timedelta import random # Only used for demo data, remove in production def generate_demo_data(model_ids, days=30): """Generate demo data for visualization purposes""" today = datetime.now() data = [] # For each model, generate 30 days of data for model_id in model_ids: base_downloads = random.randint(10, 1000) base_likes = random.randint(5, 200) # Generate daily data with some randomness and trend for i in range(days): date = (today - timedelta(days=days-i-1)).strftime("%Y-%m-%d") daily_downloads = max(1, int(base_downloads * (1 + 0.1 * i/days) * random.uniform(0.8, 1.2))) daily_likes = max(0, int(base_likes * (1 + 0.05 * i/days) * random.uniform(0.7, 1.3))) data.append({ "model_id": model_id, "date": date, "downloads": base_downloads + daily_downloads, "likes": base_likes + daily_likes, "daily_downloads": daily_downloads, "daily_likes": daily_likes }) # Update base values for next day (cumulative) base_downloads += daily_downloads base_likes += daily_likes return pd.DataFrame(data) def render_analytics_page(): st.title("📊 Model Performance Analytics") if "models" not in st.session_state or not st.session_state.models: st.info("No models found. Please create or import models first.") return # Get model IDs from the session state model_ids = [model.modelId for model in st.session_state.models] # Time period selection time_period = st.selectbox( "Select Time Period", ["Last 7 days", "Last 30 days", "Last 90 days", "All time"], index=1 ) # Convert time period to days days_mapping = { "Last 7 days": 7, "Last 30 days": 30, "Last 90 days": 90, "All time": 180 # Default to 6 months for demo } days = days_mapping[time_period] # In a real implementation, we would fetch this data from the Hugging Face API # For now, generate demo data df = generate_demo_data(model_ids, days) # Model selection for detailed view selected_models = st.multiselect( "Select Models to Compare", model_ids, default=model_ids[:min(3, len(model_ids))] ) if not selected_models: st.warning("Please select at least one model to display analytics.") return # Filter data for selected models filtered_df = df[df["model_id"].isin(selected_models)] # Create tabs for different analytics views tab1, tab2, tab3, tab4 = st.tabs(["Downloads", "Likes", "Growth Rate", "Comparison"]) with tab1: st.subheader("Downloads Over Time") # Cumulative downloads fig_cumulative = px.line( filtered_df, x="date", y="downloads", color="model_id", title="Cumulative Downloads", labels={"downloads": "Total Downloads", "date": "Date", "model_id": "Model"} ) st.plotly_chart(fig_cumulative, use_container_width=True) # Daily downloads fig_daily = px.bar( filtered_df, x="date", y="daily_downloads", color="model_id", title="Daily Downloads", labels={"daily_downloads": "Daily Downloads", "date": "Date", "model_id": "Model"}, barmode="group" ) st.plotly_chart(fig_daily, use_container_width=True) with tab2: st.subheader("Likes Over Time") # Cumulative likes fig_cumulative = px.line( filtered_df, x="date", y="likes", color="model_id", title="Cumulative Likes", labels={"likes": "Total Likes", "date": "Date", "model_id": "Model"} ) st.plotly_chart(fig_cumulative, use_container_width=True) # Daily likes fig_daily = px.bar( filtered_df, x="date", y="daily_likes", color="model_id", title="Daily Likes", labels={"daily_likes": "Daily Likes", "date": "Date", "model_id": "Model"}, barmode="group" ) st.plotly_chart(fig_daily, use_container_width=True) with tab3: st.subheader("Growth Metrics") # Calculate growth rates growth_data = [] for model in selected_models: model_data = filtered_df[filtered_df["model_id"] == model] if len(model_data) >= 2: first_day = model_data.iloc[0] last_day = model_data.iloc[-1] # Calculate download growth if first_day["downloads"] > 0: download_growth = (last_day["downloads"] - first_day["downloads"]) / first_day["downloads"] * 100 else: download_growth = 100 if last_day["downloads"] > 0 else 0 # Calculate like growth if first_day["likes"] > 0: like_growth = (last_day["likes"] - first_day["likes"]) / first_day["likes"] * 100 else: like_growth = 100 if last_day["likes"] > 0 else 0 growth_data.append({ "model_id": model, "download_growth": download_growth, "like_growth": like_growth, "downloads": last_day["downloads"], "likes": last_day["likes"] }) growth_df = pd.DataFrame(growth_data) # Show growth rates if not growth_df.empty: col1, col2 = st.columns(2) with col1: fig = px.bar( growth_df, x="model_id", y="download_growth", title="Download Growth Rate (%)", labels={"download_growth": "Growth (%)", "model_id": "Model"}, color="download_growth", color_continuous_scale=px.colors.sequential.Blues, ) st.plotly_chart(fig, use_container_width=True) with col2: fig = px.bar( growth_df, x="model_id", y="like_growth", title="Like Growth Rate (%)", labels={"like_growth": "Growth (%)", "model_id": "Model"}, color="like_growth", color_continuous_scale=px.colors.sequential.Reds, ) st.plotly_chart(fig, use_container_width=True) else: st.info("Not enough data to calculate growth rates.") with tab4: st.subheader("Model Comparison") # Get the most recent data point for each model latest_data = filtered_df.groupby("model_id").last().reset_index() # Create a radar chart for model comparison categories = ["downloads", "likes", "daily_downloads", "daily_likes"] fig = go.Figure() for model in latest_data["model_id"]: model_row = latest_data[latest_data["model_id"] == model].iloc[0] # Normalize values for radar chart (0-1 scale) max_vals = latest_data[categories].max() normalized_vals = [model_row[cat]/max_vals[cat] if max_vals[cat] > 0 else 0 for cat in categories] fig.add_trace(go.Scatterpolar( r=normalized_vals, theta=["Total Downloads", "Total Likes", "Daily Downloads", "Daily Likes"], fill='toself', name=model )) fig.update_layout( polar=dict( radialaxis=dict( visible=True, range=[0, 1] )), showlegend=True ) st.plotly_chart(fig, use_container_width=True) # Comparison table st.subheader("Numeric Comparison") comparison_df = latest_data[["model_id", "downloads", "likes", "daily_downloads", "daily_likes"]] comparison_df.columns = ["Model", "Total Downloads", "Total Likes", "Daily Downloads", "Daily Likes"] st.dataframe(comparison_df, use_container_width=True) # Analytics insights st.subheader("📈 Key Insights") # Calculate some basic insights if not filtered_df.empty: # Most downloaded model most_downloaded = filtered_df.loc[filtered_df.groupby("model_id")["downloads"].idxmax()] # Fastest growing model in terms of downloads growth_rates = [] for model in selected_models: model_data = filtered_df[filtered_df["model_id"] == model] if len(model_data) >= 2: first_downloads = model_data.iloc[0]["downloads"] last_downloads = model_data.iloc[-1]["downloads"] growth_rate = (last_downloads - first_downloads) / max(1, first_downloads) growth_rates.append((model, growth_rate)) col1, col2 = st.columns(2) with col1: st.info(f"💡 Most downloaded model: **{most_downloaded['model_id']}** with **{most_downloaded['downloads']}** total downloads") if growth_rates: fastest_growing = max(growth_rates, key=lambda x: x[1]) st.info(f"💡 Fastest growing model: **{fastest_growing[0]}** with a growth rate of **{fastest_growing[1]*100:.2f}%**") with col2: # Most liked model most_liked = filtered_df.loc[filtered_df.groupby("model_id")["likes"].idxmax()] st.info(f"💡 Most liked model: **{most_liked['model_id']}** with **{most_liked['likes']}** total likes") # Average daily downloads avg_daily = filtered_df.groupby("model_id")["daily_downloads"].mean().reset_index() highest_avg = avg_daily.loc[avg_daily["daily_downloads"].idxmax()] st.info(f"💡 Highest avg daily downloads: **{highest_avg['model_id']}** with **{highest_avg['daily_downloads']:.1f}** downloads/day")