API Latency Template

Author: amrutha97

preswald/templates/ab-testing/hello.py.template

@@ -1,107 +1,100 @@
 from preswald import (
-    Workflow, connect, get_df, text, table, plotly, selectbox,
-    slider, checkbox, separator
+    Workflow,
+    connect,
+    get_df,
+    table,
+    text,
+    plotly,
+    selectbox,
+    slider,
+    separator,
 )
-import pandas as pd
 import plotly.express as px
 
 workflow = Workflow()
 
+# --- 1. Load data and fix types ---
 @workflow.atom()
-def init():
-    text("# 🧪 A/B Testing Dashboard")
-    text("Analyze A/B test results, compare conversion rates, evaluate statistical significance, and discover top-performing variants.")
-    connect()
-
-@workflow.atom(dependencies=["init"])
 def load_data():
-    df = get_df("sample_csv").copy()
-    df["Conversion Rate"] = df["Conversion Rate"].str.replace('%', '', regex=False).astype(float)
-    df["Significance"] = df["Significance"].astype(str)
-    table(df, title="📋 Raw Data", limit=10)
-    return df
-
-@workflow.atom(dependencies=["load_data"])
-def filter_significant(load_data):
-    show_only = checkbox("Show only significant results?", default=False)
-    return load_data[load_data["Significance"].str.lower() == "yes"] if show_only else load_data
+    connect()
+    df = get_df("sample_csv")
 
-@workflow.atom(dependencies=["filter_significant"])
-def variant_selector(filter_significant):
-    variants = filter_significant["Variant"].unique().tolist()
-    return selectbox("🎛️ Choose a Variant", options=variants, default=variants[0])
+    # Coerce latency columns to float
+    df["P50"] = df["P50"].astype("float64")
+    df["P95"] = df["P95"].astype("float64")
+    df["P99"] = df["P99"].astype("float64")
 
-@workflow.atom(dependencies=["filter_significant", "variant_selector"])
-def show_variant_table(filter_significant, variant_selector):
-    df = filter_significant[filter_significant["Variant"] == variant_selector].reset_index(drop=True).copy()
-    df["Test Number"] = df.index + 1
-    table(df, title=f"🔍 Details for {variant_selector}")
     return df
 
-@workflow.atom(dependencies=["show_variant_table"])
-def plot_variant_trend(show_variant_table):
-    df = show_variant_table
-    text(f"## 📈 Conversion Trend: {df['Variant'].iloc[0]}")
-    fig = px.line(
-        df, x="Test Number", y="Conversion Rate", markers=True,
-        title="Conversion Rate Over Time", labels={"Conversion Rate": "Conversion Rate (%)"}
-    )
-    plotly(fig)
+# --- 2. Intro text ---
+@workflow.atom()
+def intro():
+    text("# 🚦 API Latency Explorer")
+    text("Explore P50, P95, and P99 latency across endpoints. Filter, chart, and inspect details.")
 
-@workflow.atom(dependencies=["filter_significant"])
-def conversion_lift_analysis(filter_significant):
-    text("## 🚀 Conversion Lift Compared to Control")
+# --- 3. Display table ---
+@workflow.atom(dependencies=["load_data"])
+def show_table(load_data):
+    text("## 📋 Full Dataset")
+    table(load_data)
 
-    # Compute average CR per variant
-    summary = filter_significant.groupby("Variant").agg(
-        AvgCR=("Conversion Rate", "mean")
-    ).reset_index()
+# --- 4. P99 threshold filter ---
+@workflow.atom(dependencies=["load_data"])
+def p99_filter(load_data):
+    text("## 🔍 Filter by P99 Latency")
+    threshold = slider("Maximum P99 (ms)", min_val=100, max_val=300, step=10, default=200)
+    filtered = load_data[load_data["P99"] <= threshold]
 
-    # Get control rate
-    control_rate = summary[summary["Variant"] == "Control"]["AvgCR"].values[0]
+    if filtered.shape[0] > 0:
+        table(filtered, title=f"Endpoints with P99 ≤ {threshold} ms")
+    else:
+        text("⚠️ No endpoints under selected threshold.")
 
-    # Calculate lift
-    summary["Lift vs Control (%)"] = summary["AvgCR"] - control_rate
+# --- 5. Plot percentiles ---
+@workflow.atom(dependencies=["load_data"])
+def plot_percentiles(load_data):
+    text("## 📊 Latency Breakdown (Grouped Bar Chart)")
+
+    melted = load_data.melt(
+        id_vars="Endpoint",
+        value_vars=["P50", "P95", "P99"],
+        var_name="Percentile",
+        value_name="Latency"
+    )
 
     fig = px.bar(
-        summary, x="Variant", y="Lift vs Control (%)",
-        color="Lift vs Control (%)", color_continuous_scale="Viridis",
-        text="Lift vs Control (%)", title="Conversion Rate Lift (vs. Control)"
+        melted,
+        x="Endpoint",
+        y="Latency",
+        color="Percentile",
+        barmode="group",
+        title="Latency by Endpoint",
+        labels={"Latency": "ms"}
     )
-    fig.update_traces(texttemplate='%{text:.2f}%', textposition='outside')
+    fig.update_layout(xaxis_tickangle=-45)
     plotly(fig)
 
-@workflow.atom(dependencies=["filter_significant"])
-def funnel_view(filter_significant):
-    text("## 🔄 Mini Funnel: Visitors → Conversions")
-    summary = filter_significant.groupby("Variant").agg({
-        "Visitors": "sum",
-        "Conversions": "sum"
-    }).reset_index()
+# --- 6. Per-endpoint breakdown ---
+@workflow.atom(dependencies=["load_data"])
+def endpoint_detail(load_data):
+    text("## 🧪 Inspect Specific Endpoint")
+    options = load_data["Endpoint"].tolist()
+    selected = selectbox("Choose Endpoint", options)
 
+    row = load_data[load_data["Endpoint"] == selected].iloc[0]
     fig = px.bar(
-        summary.melt(id_vars="Variant", value_vars=["Visitors", "Conversions"]),
-        x="Variant", y="value", color="variable", barmode="group",
-        title="Visitor vs Conversion Totals", text="value",
-        labels={"value": "Count", "variable": "Stage"}
+        x=["P50", "P95", "P99"],
+        y=[row["P50"], row["P95"], row["P99"]],
+        labels={"x": "Percentile", "y": "Latency (ms)"},
+        title=f"Latency Profile: {selected}"
     )
     plotly(fig)
 
-@workflow.atom(dependencies=["filter_significant"])
-def callout_best_variant(filter_significant):
-    summary = (
-        filter_significant.groupby("Variant")["Conversion Rate"]
-        .mean()
-        .reset_index()
-        .sort_values(by="Conversion Rate", ascending=False)
-    )
-    best = summary.iloc[0]
-    text(f"## 🏆 Best Variant: **{best['Variant']}**")
-    text(f"Achieved an average **{best['Conversion Rate']:.2f}%** conversion rate.")
-
+# --- 7. Footer ---
 @workflow.atom()
-def wrap_up():
+def end_note():
     separator()
-    text("_This interactive dashboard is built with [Preswald](https://preswald.com), empowering anyone to build production dashboards in Python._")
+    text("✅ Dashboard complete. Use this to spot latency bottlenecks in your backend!")
 
-workflow.execute()
+# --- Run ---
+workflow.execute()

preswald/templates/api-latency/sample.csv.template

@@ -27,5 +27,4 @@ Endpoint,P50,P95,P99,Last Updated
 /payments/verify,110,160,220,2022-09-15 14:45:00
 /reviews/delete,75,120,170,2022-09-15 15:00:00
 /cart/checkout,95,140,190,2022-09-15 15:15:00
-/shipping/estimate,105,160,220,2022-09-15 15:30:00
-...
+/shipping/estimate,105,160,220,2022-09-15 15:30:00