streamlit-development

Streamlit Development

Build interactive data applications using Streamlit, test them locally, and deploy to Snowflake's native Streamlit environment.

Quick Start

Execution Modes:

• Local Development - PyPI packages, full environment
• Snowflake Deployment - Snowflake Anaconda packages, managed environment

Connection Pattern

Critical: Support both local development and Snowflake deployment:

import streamlit as st
from snowflake.snowpark.context import get_active_session
from snowflake.snowpark import Session

@st.cache_resource
def get_snowpark_session():
    try:
        return get_active_session()  # Snowflake
    except:
        return Session.builder.config('connection_name', 'default').create()  # Local

For connection setup, see the snowflake-connections skill for:

• Creating ~/.snowflake/connections.toml
• Authentication methods (SSO, key pair, username/password)
• Multiple environment configurations
• Environment variable overrides

Local Development

Setup

# Install dependencies (pin Streamlit version to match Snowflake)
uv pip install --system -r requirements.txt

# Example requirements.txt
streamlit==1.46.0  # Must match Snowflake version
snowflake-snowpark-python
pandas

Run Locally

streamlit run app.py
# Or with environment overrides (see snowflake-connections skill)
SNOWFLAKE_DATABASE=MY_DB streamlit run app.py

Testing

Unit/Integration Tests

pytest streamlit_app/tests/ -v

Browser Testing

Use Playwright MCP for interactive testing:

• Verify pages load without errors
• Test forms and navigation
• Check responsive design

See TESTING_GUIDE.md for patterns.

Deployment

Option 1: Snowflake CLI (Recommended)

snow streamlit deploy --replace -c default

Option 2: Schemachange

Include Streamlit deployment in migration scripts.

Key Patterns

• Separate data access from UI logic
• Cache Snowpark sessions
• Use forms for multi-field input
• Handle errors gracefully
• Limit column nesting (max 2 levels)