I’m getting a persistent “Error installing requirements” failure on Streamlit Community Cloud. This is not a requirements.txt problem — the failure happens during the apt (system package) step, before pip ever runs, and it’s failed identically on every rebuild/reboot I’ve tried over the last day.
Log showing 4 separate attempts, same error each time (note the “invalid since” duration climbing):
[13:35:25] E: Release file for http://deb.debian.org/debian-security/dists/bullseye-security/InRelease is expired (invalid since 16h 22min 20s).
[13:39:42] E: Release file for http://deb.debian.org/debian-security/dists/bullseye-security/InRelease is expired (invalid since 16h 26min 37s).
[13:50:58] E: Release file for http://deb.debian.org/debian-security/dists/bullseye-security/InRelease is expired (invalid since 16h 37min 53s).
[14:00:19] E: Release file for http://deb.debian.org/debian-security/dists/bullseye-security/InRelease is expired (invalid since 16h 47min 14s).
Each time: installer returned a non-zero exit code / Error during processing dependencies!
What I’ve tried: Rebooting the app multiple times (log above), which just re-triggers the same failing build. As a workaround, I emptied packages.txt so the apt step is skipped entirely — that lets the deploy succeed, but I’d rather not run without tesseract-ocr (needed for OCR on scanned PDFs) long-term.
Versions: Streamlit 1.40.2, Python 3.12
Deployment: Streamlit Community Cloud (not running locally)
packages.txt (before my workaround):
tesseract-ocr
requirements.txt (full contents):
streamlit==1.40.2
pdfplumber==0.11.4
pymupdf==1.24.14
openpyxl==3.1.5
python-docx==1.2.0
anthropic==0.111.0
openai==2.42.0
python-dotenv==1.2.2
pillow==11.3.0
pandas==2.3.3
pytesseract==0.3.13
rapidfuzz==3.10.1
This looks like a stale/broken Debian security mirror baked into the build image on Streamlit’s side (bullseye-security’s Release file is never refreshing, only getting more expired with each attempt), not something fixable from my repo. Has anyone else hit this, or does the Streamlit team know about it?