Streamlit Development

9.2. Streamlit Development#

Streamlit is a Python framework purpose-built for creating data science and machine learning applications. Unlike Voila (which converts notebooks), Streamlit apps are Python scripts designed from the ground up for interactive web applications.

9.2.1. Why Streamlit?#

  • Pure Python: No HTML/CSS/JavaScript required

  • Rapid development: Build apps in hours

  • Built for ML: Optimized for data science workflows

  • Active development: Rich ecosystem and community

  • Easy deployment: Streamlit Cloud (free hosting)

9.2.2. Installation#

pip install streamlit

9.2.3. Your First Streamlit App#

Create app.py:

import streamlit as st

st.title("My First Streamlit App")

name = st.text_input("Enter your name:")

if name:
    st.write(f"Hello, {name}!")

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
Cell In[1], line 1
----> 1 import streamlit as st
      3 st.title("My First Streamlit App")
      5 name = st.text_input("Enter your name:")

ModuleNotFoundError: No module named 'streamlit'

Run:

streamlit run app.py

Opens automatically at http://localhost:8501

9.2.4. Streamlit Basics#

9.2.4.1. Display Content#

import streamlit as st

# Text
st.title("Title")
st.header("Header")
st.subheader("Subheader")
st.text("Plain text")
st.markdown("**Bold** and *italic*")
st.write("Anything:", data, object, plot)

# Code
st.code("print('Hello')", language='python')

# LaTeX
st.latex(r"e^{i\\pi} + 1 = 0")

9.2.4.2. Input Widgets#

# Text input
name = st.text_input("Name:")
email = st.text_input("Email:", type="password")
text = st.text_area("Comments:")

# Numbers
age = st.number_input("Age:", min_value=0, max_value=120, value=25)
slider_val = st.slider("Select value:", 0, 100, 50)

# Selection
option = st.selectbox("Choose one:", ["A", "B", "C"])
multi = st.multiselect("Choose multiple:", ["X", "Y", "Z"])
checked = st.checkbox("Agree to terms")
radio = st.radio("Select:", ["Option 1", "Option 2"])

# Buttons
if st.button("Click me"):
    st.write("Button clicked!")

# File upload
uploaded = st.file_uploader("Choose a CSV", type="csv")

# Date/Time
date = st.date_input("Select date")
time = st.time_input("Select time")

# Color picker
color = st.color_picker("Pick a color", "#00f900")

9.2.4.3. Layout#

Columns:

col1, col2, col3 = st.columns(3)

with col1:
    st.header("Column 1")
    st.write("Content for column 1")

with col2:
    st.header("Column 2")
    st.write("Content for column 2")

with col3:
    st.header("Column 3")
    st.write("Content for column 3")

Expander:

with st.expander("Click to expand"):
    st.write("Hidden content")
    st.image("image.png")

Sidebar:

st.sidebar.title("Settings")
threshold = st.sidebar.slider("Threshold:", 0.0, 1.0, 0.5)
model_type = st.sidebar.selectbox("Model:", ["Linear", "RF", "GB"])

Tabs:

tab1, tab2, tab3 = st.tabs(["Data", "Model", "Results"])

with tab1:
    st.header("Data Overview")
    st.dataframe(df)

with tab2:
    st.header("Model Configuration")
    # Model settings

with tab3:
    st.header("Results")
    # Show results

9.2.5. ML Model Application#

import streamlit as st
import joblib
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

st.title("🔮 Customer Churn Prediction")

# Load model
@st.cache_resource
def load_model():
    return joblib.load('models/churn_model.joblib')

model = load_model()

# Sidebar for inputs
st.sidebar.header("Customer Information")

age = st.sidebar.slider("Age", 18, 80, 35)
income = st.sidebar.slider("Annual Income ($)", 20000, 200000, 50000, step=5000)
credit_score = st.sidebar.slider("Credit Score", 300, 850, 650)

# Feature engineering
features = pd.DataFrame({
    'age': [age],
    'income': [income],
    'credit_score': [credit_score]
})

# Main area
st.header("Customer Profile")
col1, col2, col3 = st.columns(3)
col1.metric("Age", age)
col2.metric("Income", f"${income:,}")
col3.metric("Credit Score", credit_score)

# Predict
if st.button("Predict Churn", type="primary"):
    prediction = model.predict(features)[0]
    probability = model.predict_proba(features)[0]
    
    st.header("Prediction")
    
    if prediction == 1:
        st.error("⚠️ HIGH RISK: Customer likely to churn")
    else:
        st.success("✅ LOW RISK: Customer likely to stay")
    
    # Probability gauge
    churn_prob = probability[1]
    st.write(f"**Churn Probability:** {churn_prob:.1%}")
    st.progress(churn_prob)
    
    # Visualization
    fig, ax = plt.subplots()
    ax.barh(['Retain', 'Churn'], probability)
    ax.set_xlabel('Probability')
    ax.set_xlim(0, 1)
    st.pyplot(fig)

9.2.6. Caching for Performance#

@st.cache_data: Cache data (DataFrames, lists, etc.)

@st.cache_data
def load_data():
    # Expensive data loading
    return pd.read_csv('large_file.csv')

df = load_data()  # Runs once, then cached

@st.cache_resource: Cache models, connections

@st.cache_resource
def load_model():
    return joblib.load('model.joblib')

model = load_model()  # Cached

9.2.7. Session State#

Maintain state across reruns:

# Initialize
if 'count' not in st.session_state:
    st.session_state.count = 0

# Update
if st.button("Increment"):
    st.session_state.count += 1

st.write(f"Count: {st.session_state.count}")

9.2.8. Deployment#

9.2.8.1. Streamlit Cloud (Free)#

  1. Push code to GitHub

  2. Go to share.streamlit.io

  3. Connect repository

  4. Deploy!

requirements.txt:

streamlit
scikit-learn
joblib
pandas
matplotlib

9.2.8.2. Docker Deployment#

Dockerfile:

FROM python:3.11-slim

WORKDIR /app

COPY requirements.txt .
RUN pip install -r requirements.txt

COPY . .

EXPOSE 8501

CMD ["streamlit", "run", "app.py", "--server.address", "0.0.0.0"]

docker build -t streamlit-app .
docker run -p 8501:8501 streamlit-app

9.2.9. Pros and Cons#

9.2.9.1. Advantages#

  • Pure Python (no web dev skills needed)

  • Rapid development

  • Built-in widgets and layouts

  • Free hosting (Streamlit Cloud)

  • Active community

9.2.9.2. Limitations#

  • Limited customization of UI

  • Reruns entire script on interaction

  • Not ideal for complex applications

  • Some performance constraints

9.2.10. When to Use Streamlit#

Use Streamlit when:

  • Building ML/data apps from scratch

  • Need more customization than Voila

  • Want rapid prototyping

  • Target is data scientists/analysts

Skip Streamlit when:

  • Need pixel-perfect custom UI

  • Building complex multi-page applications

  • Need fine-grained control over frontend

9.2.11. Summary#

Streamlit provides:

  • Fast development of data/ML apps

  • Pure Python interface

  • Built-in widgets and layouts

  • Easy deployment to Streamlit Cloud

  • Good balance between simplicity and functionality

Contents