Daily digest of news/media mentions of Jodie Rummer, RummerLab, Physioshark, and Athletes of the Reef. Complements Google Alerts and Talkwalker Alerts. It does not scrape full article pages.
Each morning it fetches targeted RSS/API sources, dedupes them, and emails new items to athletesofthereef@gmail.com in a Google Alerts-style message.
Search query
"Jodie Rummer" OR RummerLab OR "Rummer Lab" OR Physioshark OR "Athletes of the Reef"
Result type: start with News only.
That matches Google Alerts and this digest. Add a second Talkwalker alert for Blogs if you want Cosmos/Oceanographic-style sites. Skip Twitter and Discussions unless you want social noise.
| Source | Notes |
|---|---|
| Google News RSS (AU) | Targeted keyword search |
| Google News RSS (US) | Catches Independent, Popular Mechanics, Forbes, etc. |
| Google News RSS (UK) | Guardian / Independent UK editions |
| Bing News RSS | Second crawler |
| The Conversation author Atom | Jodie L. Rummer |
| GDELT DOC 2.0 | Free global article search; no API key |
| Guardian Open Platform | Optional THE_GUARDIAN_API_KEY |
| NewsAPI.org | Optional NEWS_API_ORG_KEY (free tier: /v2/top-headlines) |
| newsapi.ai / Google CSE / etc. | Keys in .env.example; sources not wired yet |
Generic outlet firehoses (whole ABC / SMH feeds) from the old Scholar scraper are not included. Those created more noise than Google Alerts. This tool only uses search feeds and APIs already scoped to Jodie/the lab.
Own sites (rummerlab.com, jodierummer.com, physioshark.org) are dropped.
- Copy
.env.exampleto.env. - Create a Gmail App Password for
athletesofthereef@gmail.comand setSMTP_PASSWORD. - Optional: register a free Guardian Open Platform key and/or NewsAPI key.
- Run:
cp .env.example .env
docker compose pull
docker compose up -d --no-buildDefault schedule: 07:00 Australia/Brisbane. Set RUN_ON_START=1 for a first send on container start.
Manual run without waiting:
docker compose run --rm digest python -m src --once
docker compose run --rm digest python -m src --dry-runSeen URLs are stored in data/seen.json so repeats are not emailed again.
The container runs as a non-root user (APP_UID/APP_GID, default 1000:1000), and that user must be able to write ./data. If your host user isn't 1000, set APP_UID/APP_GID in .env to the output of id -u / id -g. Images up to v0.0.2 ran as root, so an existing data/ may be root-owned. Fix it once before upgrading:
sudo chown -R "${APP_UID:-1000}:${APP_GID:-1000}" datapython -m venv .venv
.venv/Scripts/activate # Windows
pip install -r requirements-dev.txt
ruff check . --fix && ruff format .
pytestProduction Compose pulls GHCR images with native ARM64 (Odroid) and AMD64 support. After merging this change and publishing the first release, run from the existing deployment directory:
git pull --ff-only
docker compose pull
docker compose up -d --no-build --remove-orphans
docker compose psKeep the same Compose project name and existing volumes/bind mounts. Keep your .env and credentials; do not delete production data. The existing Watchtower installation can follow latest using the containers' com.centurylinklabs.watchtower.enable=true labels. If Watchtower filters names or scopes, include these containers in that configuration. No second Watchtower instance is needed. Watchtower applies new images, while future Compose and environment changes still need git pull and docker compose up. GHCR packages must be public for anonymous pulls; alternatively configure registry authentication in Docker and Watchtower.
Set IMAGE_TAG in .env to a published vMAJOR.MINOR.PATCH tag to pin the digest image or roll back. Local builds use docker compose -f docker-compose.yml -f docker-compose.build.yml up -d --build.
docker-release.yml publishes stable v* tag pushes, manual runs with an existing version tag, and calls from the nightly updater. It checks the exact release commit, validates both native architectures without secrets or upstream services, publishes the GitHub release, then updates the multi-platform latest tag only if no newer stable release has been published. Rerunning an older release leaves latest unchanged. To publish manually, push an unused stable version tag on main, or rerun the Docker release workflow with that existing tag. The normal GITHUB_TOKEN needs contents/packages write permission; no PAT is required. Tag publication created by GITHUB_TOKEN does not trigger other workflows, so nightly calls the reusable release workflow explicitly.
Nightly updates run at 00:37 Australia/Brisbane. Python direct constraints live in requirements.txt and requirements-dev.txt; their generated requirements.lock and requirements-dev.lock pin the full graph with hashes. uv excludes versions uploaded in the last three days and keeps upgrades within each locked package's major (within its minor for 0.x), without downgrading. The age gate reduces exposure to newly compromised releases; it does not guarantee package safety.
Only a changed production Python graph creates a patch version, tag, GitHub release, and image. This includes transitive and platform-specific runtime changes. Developer tooling and hash-only lock changes are committed after checks but do not release. With no changes, the run does nothing. Tests, lint, and a local production image smoke test run before a production update is committed; the release then checks both architectures. The bot refuses to rebase if main changed during validation, and pushes the tested commit/tag atomically. Branch rules must allow this bot to push dependency commits to main; failures remain visible in Actions rather than bypassing protections.
To refresh manually, install uv==0.13.7, then run python scripts/dependency_release.py update. Do not edit generated lockfiles by hand. Nightly uses existing stable tags when choosing the next patch version, even if project metadata previously lagged behind a release.