For Developers

Want to build something like this? Here's how we indexed 109,000+ documents from the DOJ Epstein Files release.

The EpsteIN Tool

EpsteIN by cfinke is the extraction tool that made this possible. It handles the DOJ's paginated PDF releases and extracts text content.

What We Did

Downloaded all 12 DOJ datasets (3M+ pages)
Ran OCR on scanned documents using Tesseract
Extracted text from native PDFs with pdftotext
Indexed everything into Meilisearch for full-text search
Added entity extraction (names, locations, dates)

API Access

Direct search against Meilisearch. No auth required for read operations.

curl -X POST "https://epstein.dugganusa.com/indexes/epstein_files/search" \
  -H "Content-Type: application/json" \
  -d '{"q": "Prince Andrew", "limit": 20}'

Response Format

{
  "hits": [
    {
      "efta_id": "EFTA00022136",
      "content": "...",
      "dataset": "dataset3",
      "pages": 5,
      "people": ["prince_andrew", "virginia_giuffre"],
      "locations": ["new_york"]
    }
  ],
  "estimatedTotalHits": 228,
  "processingTimeMs": 12
}

Breaking eggs: This index exists because someone decided completeness mattered more than perfect formatting. If you're building something similar, don't let perfect be the enemy of done. Index what you have, improve later.

Data Completeness

We indexed:

Datasets 1-12 from the DOJ SDNY release
~109,000 documents
Both native text and OCR'd scans
Entity extraction for names/places

Want the Raw Data?

The source documents are public record from justice.gov. Our index adds searchability and entity extraction on top.

Questions? Contact us.