[self-promotion] Full Historical GDELT Event Archive (50+ Years, Hundreds Of Millions Of Rows) As Clean, Sampled Parquet, No API Or BigQuery Limits

Disclosure: I’m the developer of the tool linked below.

GDELT (the Global Database of Events, Language, and Tone) is one of the largest open event datasets around: every reported event worldwide since 1979, tagged by actor, location, tone, and (via the Global Knowledge Graph) the article that reported it. The catch is actually getting the full thing: the official API caps you at ~250 rows per query and a 3-month window, and BigQuery’s free tier (1TB/month query, 10GB storage) doesn’t stretch to a full historical pull either. The raw bulk archive at data.gdeltproject.org is open, but it’s thousands of loose CSV/ZIP files with no sampling or joining logic of its own.

I built GdeltForge to close that gap: it downloads, checksums, converts to Parquet, and reproducibly samples the whole archive locally, and can cross-reference a sample back onto the Global Knowledge Graph. Check out the PyPI page, the GitHub Repo, the documentation, and especially the hands-on getting started notebook, runnable directly in Colab, no install needed. You can generate you own samples or even scan the entire data to create a dataset