Building the JFK Files Explorer: Making History Accessible with Vectorize

Chris Bartholomew•
Building the JFK Files Explorer: Making History Accessible with Vectorize

With the release of the latest JFK assassination files this week, we saw an opportunity to showcase Vectorize’s capabilities while providing a valuable service to the public. Today, I’m excited to share JFK Files Explorer, a free tool that allows anyone to have conversations with these historically significant documents.

The Challenge: Scale and Complexity

The JFK files release presented a perfect test case for our technology:

  • Massive Scale: Over 2,000 documents spanning 65,000 pages
  • Complex Document Formats: Primarily scanned documents, many with poor quality
  • Historical Significance: Documents of intense public interest requiring accurate extraction
  • Quick Turnaround: Need to make the files accessible while they’re still newsworthy

Most document intelligence platforms would struggle with the volume and quality of these files. Traditional OCR often fails on low-quality scans, while manual processing would take weeks or months.

Vectorize: Built for This Challenge

Our platform was perfectly positioned to tackle this project:

  1. Vectorize Iris Extraction Model: Our specialized model excels at extracting text from challenging document formats, including the low-quality scans prevalent in the JFK collection
  2. Scalable Processing Pipeline: Vectorize processed all 65,000 pages in hours rather than days or weeks
  3. Built-in Vector Database: Efficiently stored and indexed the extracted content for semantic search
  4. Retrieval Pipeline: Our RAG implementation ensures relevant documents are retrieved based on semantic similarity

Technical Stack: How We Built It

For those interested in the technical details, here’s how we built the JFK Files Explorer:

  • Data Processing: Vectorize Iris extraction model for document text extraction
  • Vector Database: Vectorize’s built-in vector database for storage and retrieval
  • Retrieval: Vectorize retrieval pipeline for semantic search capabilities
  • Frontend: Next.js application using Vercel’s AI SDK
  • LLM: Llama 3.3 70B Versatile hosted on Groq
  • Deployment: Containerized with Docker and deployed on Kubernetes

The entire application is powered by Vectorize RAG-as-a-Service, demonstrating how our platform enables rapid development of sophisticated document intelligence applications.

Making History Accessible

Beyond showcasing our technology, this project serves an important purpose: democratizing access to historical documents. By providing a conversational interface to the JFK files, we’ve made these documents accessible to everyone—not just researchers with specialized tools or the patience to manually search through thousands of pages.

Users can ask questions in natural language and receive contextually relevant answers with citations back to the original documents. This transforms the research process from tedious document scanning to an intuitive conversation.

Try It Yourself

We’ve made the JFK Files Explorer completely free for anyone to use. Whether you’re a historian, researcher, journalist, or simply curious about this chapter in American history, you can start exploring the files today.

This project represents what we’re passionate about at Vectorize—using advanced document intelligence to make information more accessible. It’s a small demonstration of how our technology can transform how we interact with documents, particularly at scale.

I’d love to hear your feedback on the tool or answer any questions about how we built it. And if you’re working with document challenges in your organization, this is exactly the kind of problem Vectorize was built to solve.

The information provided in the JFK Files Explorer is derived directly from public documents released by the U.S. National Archives and is provided without any warranty or guarantee of accuracy.