ReadMeFeed / Community knowledge
Real questions. Useful conversations. Find the people who know your stack.
Developer tools · #streamlit
Streamlit’s @st.cache_data memoizes heavy data functions by hashing arguments, reducing recomputation on reruns. This guide shows how to configure TTL, size limits, and avoid stale data with a concrete CSV‑loading example.
Streamlit's st.fragment decorator isolates UI sections so only that fragment re-runs on widget changes — cutting dashboard latency from seconds to milliseconds without replacing caching.
Learn how to preserve widget values and app logic across Streamlit script reruns using session_state, with initialization, binding, reading, and reset patterns.
Learn how @st.fragment limits reruns to UI-only changes while preserving @st.cache_data and @st.cache_resource, with a practical example and notes on limits.
Learn how to use Streamlit’s st.cache_data decorator to avoid re‑loading large datasets on every interaction, with a concrete CSV‑loading example, version notes, and cache‑staleness mitigation.
Learn how to decide when to cache a Streamlit function, what boundaries the cache respects, and how to verify that caching improves performance without introducing stale data.
Learn when to use @st.cache_data and @st.cache_resource in Streamlit, see a worked example, and understand the trade‑offs of cached values.
Learn when and how to use Streamlit's st.cache_data decorator, including purity and hashability requirements, memory boundaries, operational checks, failure modes, and triggers for redesign.
Streamlit 1.18 split caching into st.cache_data (serializable returns) and st.cache_resource (singleton objects). Covers when to use each, TTL/LRU eviction, widget pitfalls, mutation risks, and distributed deployment patterns with a semantic search example.
python