Theodore Roosevelt Presidential Library Labs

Built here.
Yours to fork.

The Theodore Roosevelt Presidential Library builds a lot of its own software — ticketing widgets, embeddable timelines, a link checker, a 3D campus map, a night-sky explorer. All of it is public, all of it is MIT licensed, and most of it runs on nothing but GitHub Pages. A small museum with no engineering staff can fork any of these, swap one data file, and have it live on their own domain the same afternoon.

48Public projects
47Live demos
MITEvery repository
$0Hosting, in most cases

The Library's AI work, published in full

The Living Library

Transforming Archival Collections into Conversational Knowledge Systems — Lessons from the Theodore Roosevelt Presidential Library

Every museum and library is being asked what it is doing about AI. This is the Library's answer, written down with Microsoft's AI for Good Lab and published for anyone to copy: how roughly 300,000 archival records became something a visitor can question in plain language, powering Campfire, the Archivist App, and the Talk to TR exhibit.

It is four layers, and the AI avatar is only the last one — and it is optional. The first three are the part most institutions actually need. It also states plainly what the authors could not resolve, which is rarer than it should be.

And the code is now public. Microsoft has released the reference implementation at github.com/microsoft/TRPL — all four layers, MIT licensed, with synthetic sample data and a local Docker setup, so a curious institution can run it before committing to anything.

Read the breakdown Paper on arXiv
One letter, end to end: Roosevelt to his son Kermit, read by OCR, then placed in the graph that makes the collection searchable. Animation: Microsoft AI for Good Lab

The four layers

  1. 4Conversational interface Talk to TROptional
  2. 3Retrieval and reasoning Campfire
  3. 2AI-powered processing Archivist App
  4. 1Digitization and corpus Theodore Roosevelt Center · Terentia

48 projects

Why any of this is public

Museums and nonprofits keep solving the same problems separately and expensively. A timed-entry sell-out warning, a photo gallery that does not cost a monthly subscription, a way to get hours out of one CMS and onto another site — these are not competitive advantages. They are plumbing. Publishing the plumbing costs the Library nothing and saves somebody else a procurement cycle.

Built to be handed off

Nearly every project keeps its content in a plain JSON or Markdown file, separate from its code. Adapting one usually means rewriting data and swapping a logo, not learning a framework. Each project page lists exactly which files to change.

Hosting is usually free

Most of these run entirely on GitHub Pages with a scheduled GitHub Action doing whatever work a server would normally do. No hosting bill, no database, no credentials sitting in a browser. Where an API key is needed, it lives in a repository secret.

Fork it, do not ask

MIT means you can use, change, and redistribute any of this commercially, without attribution and without asking. Open an issue if something is broken or unclear — that feedback makes the next fork easier for somebody else.

Worth saying plainly: these are not products. There is no support contract and no roadmap, they were built for one institution's circumstances, and a few are prototypes that say so on their own page. Licensed fonts and collection photography are not the Library's to redistribute, so every project falls back to open substitutes.

About the Library