Streamlit st.cache_data and Docker health‑check integration: does caching persist across container stop/start?

Answer

By default, st.cache_data stores objects only in the Python process memory. When the Docker container stops, the process ends and the in‑memory cache is discarded, so a health‑check‑triggered restart does not preserve the cache.

Confirmed facts

• Streamlit ≥1.18 adds a persist argument to st.cache_data (True or \"disk\") that writes cached values to the local filesystem under ~/.streamlit.
• If that directory is bind‑mounted or declared as a Docker volume, the persisted cache survives container stop/start and is reused on subsequent runs.
• Without persist=True the only way to retain data across restarts is to store it externally (e.g., a database, Redis, or object store) and reload it inside the cached function.

Steps to enable persistence for low‑traffic workloads

1. Verify you are using Streamlit ≥ 1.18 (see diagnostic note below).
2. Annotate the caching function with @st.cache_data(persist=True).
3. Create a host directory for the cache, e.g., mkdir -p ./streamlit_cache.
4. Run the container with a volume mount pointing to the Streamlit cache directory:

   docker run -d -v $(pwd)/streamlit_cache:/home/streamuser/.streamlit -p 8501:8501 your-streamlit-image

5. After the first request, the cached value will be written to the mounted directory. Subsequent container starts (triggered by the health‑check) will read the same file, giving you persistence without changing the caching logic.

Missing diagnostic detail

To confirm that the persist argument is available, please share the exact Streamlit version you are running (e.g., pip show streamlit). If you are on an older version, the persist option will be ignored and you will need an external store instead.