optimizing-streamlit-performance

To speed up Streamlit apps, apply caching decorators like @st.cache_data for datasets and @st.cache_resource for models. This persists expensive objects across reruns, preventing heavy computations from executing on every user interaction.

Choose st.cache_data for caching data structures like DataFrames, and st.cache_resource for global resources like database connections or models. Using the correct decorator ensures heavy objects persist efficiently across reruns without memory duplication.

Streamlit fragments isolate reruns to specific UI components, while forms batch user inputs to delay execution. Combining st.fragment and st.form prevents the main script from rerunning entirely, minimizing unnecessary recomputation for interactive dashboards.

Yes, you can optimize Streamlit widgets by moving costly computations out of the main flow and loading data only as needed. Pre-computation and lazy loading improve responsiveness for interactive tools rendering heavy UI components.

A Streamlit dashboard runs slow if caching is misapplied or heavy UI components trigger full reruns. Isolate widget reruns with fragments, batch inputs with forms, and ensure pre-computation moves heavy logic out of the main execution flow.

Avoid st.cache_data for mutable global resources like database connections or machine learning models. Use st.cache_resource instead to persist these objects, preventing the overhead of re-initializing expensive connections on every app rerun.