How to host a Streamlit app online for free

Streamlit is the fastest way to turn a machine-learning model into a web app. Getting it online is where most students get stuck. Here's exactly how to do it, using a real house-price predictor as the example.

Live example: houseprice-predictor.shipzo.tech, a Streamlit + scikit-learn app deployed straight from GitHub with the steps below.

Step 1: Structure your repository

house-price-predictor/
├── app.py              # your Streamlit app
├── model.pkl           # trained model (or train it at start-up)
├── data.csv            # any data files the app reads
├── requirements.txt
└── tests/
    └── test_model.py   # optional: runs before every deploy

Keep app.py in the root folder. If your app lives in a sub-folder, you can point the host to it later with a root directory setting.

Step 2: Pin your requirements

requirements.txt

streamlit>=1.39
pandas>=2.2
scikit-learn>=1.5
joblib>=1.4

Pin the scikit-learn version you trained with. A model pickled with one version can fail to load, or give different results, on another.

Step 3: Load files and models the right way

The most common Streamlit deploy error is FileNotFoundError. Paths that work on your laptop break on a server because the working directory is different. Build paths from the script's own location, and cache the model so it loads once instead of on every click:

app.py

from pathlib import Path
import joblib
import pandas as pd
import streamlit as st

BASE = Path(__file__).parent

@st.cache_resource
def load_model():
    return joblib.load(BASE / "model.pkl")

@st.cache_data
def load_data():
    return pd.read_csv(BASE / "data.csv")

st.title("🏠 House Price Predictor")
area = st.number_input("Area (sq ft)", 300, 10000, 1100, step=50)
bedrooms = st.slider("Bedrooms", 1, 6, 2)

if st.button("Predict"):
    price = load_model().predict([[area, bedrooms]])[0]
    st.metric("Estimated price", f"₹ {price:,.1f} lakh")

Step 4: Push to GitHub

git init
git add .
git commit -m "Streamlit app"
git branch -M main
git remote add origin https://github.com/YOUR-USERNAME/house-price-predictor.git
git push -u origin main

Step 5: Deploy

1. Sign up on Shipzo (free) and open New deployment.
2. Paste the repo link.
3. Choose Standard if you use scikit-learn, TensorFlow or PyTorch, or Starter for light apps. Then press Deploy.

Shipzo spots streamlit in your requirements and runs the equivalent of:

streamlit run app.py --server.port $PORT --server.address 0.0.0.0 --server.headless true

Streamlit talks to the browser over a websocket, which some proxies block (you'd see the app stuck on "Please wait…"). Shipzo passes websockets through, so this just works.

The first build of a data-science app takes 5–8 minutes because pandas and scikit-learn are large. Later deploys are faster.

Optional: test your model before every deploy

Add a quick pytest file and every push is checked before it goes live:

tests/test_model.py

from pathlib import Path
import joblib

MODEL = joblib.load(Path(__file__).parent.parent / "model.pkl")

def test_price_is_positive():
    assert MODEL.predict([[1000, 2]])[0] > 0

def test_more_area_costs_more():
    assert MODEL.predict([[2000, 2]])[0] > MODEL.predict([[1000, 2]])[0]

Common errors and fixes

Frequently asked questions