St.download_button working intermittently on cloud

Using Streamlit
,
1

I am using st.download_button to download generated pdf files and xls files.

Here is the sample code:

st.download_button(
label=“PDF Report”,
data=decoded_pdf_attachment,
file_name= idea_id+“_Analysis_Report.pdf”,
mime=“application/pdf”
)

The decoded_pdf_attachment is an io.BytesIO string. The button works perfectly on my local machine (localhost). However, when I deploy in the cloud, it works intermittently. Sometimes it downloads a pdf file but most of the time it can’t find a file and downloads an html like this:
e7a7b77253b047af63140c3ae6bd82c4a64880dc38597c7e3d3e65f4.pdf.html
With the content
404: Not Found

Can this be fixed and made more consistent. I am using version 1.45 of streamlit.

Thanks

Sudipta Bhawmik

2

Welcome to the Streamlit community, Sudipta! Thanks for sharing your question and code snippet—this is a common issue when moving from local to cloud deployments. The 404 error and the download of an unexpected HTML file usually indicate that the file-like object (your io.BytesIO) is either empty, closed, or not available at the time the download button is rendered in the cloud environment.

To ensure consistent downloads, make sure:

  • The io.BytesIO object is open, contains the correct data, and its pointer is reset to the beginning (use decoded_pdf_attachment.seek(0) before passing it to st.download_button).
  • The data is generated and available in memory before the button is rendered, as Streamlit expects the data to be ready at render time.
  • You’re not relying on files written to disk, as cloud environments may not persist files between reruns or sessions.

Here’s a robust pattern:

import streamlit as st
import io

# Example: create a PDF in memory (replace with your actual PDF generation)
pdf_bytes = b"%PDF-1.4..."  # Your PDF bytes here
pdf_buffer = io.BytesIO(pdf_bytes)
pdf_buffer.seek(0)

st.download_button(
    label="PDF Report",
    data=pdf_buffer,
    file_name=f"{idea_id}_Analysis_Report.pdf",
    mime="application/pdf"
)

If you generate the PDF dynamically, ensure the buffer is not closed or garbage-collected before the button is clicked. For large or expensive-to-generate files, consider using a callable for the data parameter (supported in recent Streamlit versions) to defer generation until the button is clicked. See the Streamlit docs for st.download_button for more details and examples.

If you’re still seeing intermittent issues, please share a minimal reproducible example and details about your cloud deployment (platform, file generation logic, etc.) so the community can help debug further. Also, check for any errors in your app logs.

Sources: