libreblog

Blog for LibreCourseUY, built with Hugo and the PaperMod theme, and auto-deployed via GitHub Actions. Every push to main builds a Docker image and publishes it to GHCR.

Local development

git clone --recurse-submodules https://github.com/LibreCourseUY/libreblog
cd libreblog
hugo server

The site is generated from content/ and configured in hugo.yaml.

Weekly TL;DR automation

A GitHub Actions workflow (.github/workflows/weekly-tldr.yml) runs every Monday at 12:00 UTC. It:

1. pulls the last week of tech news from the RSS feeds in scripts/feeds.txt;
2. asks Gemini to write an edition in the house style (scripts/tldr_style.md, using the latest editions as examples) with exactly 14 news items;
3. validates the result (front matter, sections, source links, footer, no em dashes, no double hyphens) and writes content/tldr/vX.Y.md;
4. opens a pull request. Merge it to publish.

Setup

• Add the repository secret GEMINI_API_KEY.
• Enable Settings -> Actions -> "Allow GitHub Actions to create and approve pull requests".

Running it by hand

python -m venv .venv
.venv/bin/pip install -r scripts/requirements.txt

# print a draft without writing or opening a PR
GEMINI_API_KEY=... .venv/bin/python scripts/generate_tldr.py --dry-run

# write the file locally (no commit, no PR)
GEMINI_API_KEY=... .venv/bin/python scripts/generate_tldr.py

Useful flags: --since-days N, --model gemini-flash-lite-latest, --force.

The generator retries transient Gemini errors and falls back to other available models when the requested one is overloaded. Set TLDR_MODELS to a comma-separated list to control the fallback order.

The workflow also supports manual dispatch with the dry_run and since_days inputs.

Tests

.venv/bin/python -m unittest scripts.test_generate_tldr

License

MIT, see LICENSE.