Tech With Galvan

Introduction

A Streamlit app running on localhost is a demo; deployed, it's a product you can put on your resume, share with clients, or submit as a portfolio piece. Streamlit Community Cloud hosts your apps free, straight from a GitHub repo, and deployment takes about five minutes once your repo is shaped correctly.

This guide covers the whole path: repo prep, deployment, secrets, and the three errors that account for most failed first deployments.

Step 1: Shape Your Repo

Your GitHub repo needs exactly three things (see the Git guide if the repo doesn't exist yet):

Your app file — e.g. app.py.

`requirements.txt` — every import, with versions:

code

pip freeze > requirements.txt

Any data files your app reads — or code that downloads them.

The requirements file is where 90% of deploy failures originate. The cloud machine starts empty; if you import plotly and it's not in requirements.txt, the app crashes with ModuleNotFoundError — locally it worked because you installed it months ago and forgot.

Step 2: Deploy

Go to share.streamlit.io and sign in with GitHub.

Click Create app → Paste repo URL.

Select your repo, branch, and the main file path (app.py).

Click Deploy.

The build log streams live: pip installing your requirements, then booting the app. Two minutes later your app lives at https://yourname-app-name.streamlit.app. Every git push to the linked branch redeploys automatically — the same push-to-deploy rhythm as the Cloudflare Pages workflow.

Step 3: Secrets the Right Way

Apps with API keys — the Mistral chatbot, the PDF chat — must not commit keys to GitHub. Streamlit's answer is secrets management:

App settings → Secrets → edit.

Add TOML-formatted secrets:

code

MISTRAL_API_KEY = "sk-..."

Read them in code:

code

import streamlit as st
key = st.secrets["MISTRAL_API_KEY"]

Locally, the same code reads a .streamlit/secrets.toml file in your project — one code path, two environments, zero committed secrets.

The Three Classic Deploy Errors

`ModuleNotFoundError` — a missing requirements entry. Add it, push, the app redeploys. Check for indirect imports too: streamlit-audio-recorder pip-installs differently from its import name.

App builds but shows a blank/error screen — usually a file-not-found: the app opens data.csv with a relative path, but the cloud's working directory differs. Use pathlib.Path(__file__).parent / "data.csv" to anchor paths to the script.

App sleeps or resets constantly — free-tier apps sleep after inactivity and restart on a fresh machine, wiping session state *and files written at runtime*. Persist to external storage (a database, S3) if data must survive.

Resource Limits to Know

The free tier gives each app ~1GB RAM and modest CPU — plenty for every tutorial on this site, tight for the image classifier's TensorFlow models (use MobileNet, not ResNet) and the larger Whisper sizes (stick to tiny/base).

Key Concepts

Push-to-deploy — the repo is the deployment trigger.

requirements.txt is the machine — the cloud builds exactly what you list.

Secrets, never env vars in code — platform-managed key storage.

Ephemeral filesystem — runtime files vanish on restart.

What to Try Next

Deploy the expense tracker — then solve its CSV persistence problem with a cloud database.

Add multi-page apps with st.navigation for a portfolio of tools in one deploy.

Put your apps behind your own domain — pair with the Cloudflare Pages guide for a unified personal site.

Monitor app health with Streamlit's built-in status dashboard.

FAQ

Is Streamlit Community Cloud really free?

Yes — unlimited apps with resource limits per app, supported by Streamlit (Snowflake). For higher limits and org features, paid tiers exist, but most learning projects never hit them.

Can I deploy private-repo apps?

Yes — authorize the Streamlit GitHub app for the repo during deployment; the flow is identical.

How do I update a deployed app?

Push to the linked branch — the app rebuilds automatically. No manual redeploy step exists, which is the feature.