Building a Searchable Database of U.S. Presidential Pardons

The Breakthrough

Vid Luther created Pardonned.com, a searchable, daily-updated database that aggregates every federal pardon and commutation since 1993, allowing users to filter by administration, crime type, and restitution amount.

Technical Implementation

The project uses a minimalist, static-site architecture designed for long-term maintenance and verifiability:

• Scraping Pipeline: A Playwright script runs in a headless browser to scrape raw DOJ clemency pages, using multiple custom parsers to handle the inconsistent formatting across different presidential administrations.
• Data Storage: All scraped data is normalized and stored in a local SQLite file, avoiding the need for a persistent database server.
• Static Generation: The site is built using Astro, which reads the SQLite file at build time to generate static pages for every individual pardon.
• Automation: A GitHub Action triggers the entire pipeline every morning, rescraping the DOJ source, rebuilding the dataset, and redeploying the site to Cloudflare Pages.
• Verifiability: The entire project is open-source, allowing users to clone the repository and run the full scraping pipeline locally to verify the data against the original government source documents.

Context

The project addresses the lack of accessibility in government-provided clemency data. While the Department of Justice publishes pardon records, they are buried in raw HTML and scanned PDFs, making cross-administration analysis or large-scale searching impossible for the public. By centralizing this data and providing a searchable interface, the tool surfaces trends, such as the fact that two out of every three pardons are for drug offenses, and allows for direct comparison of restitution amounts forgiven by different administrations.