Ben Welsh🔍 I'm pleased to announce the launch of archive.ire.org, a new AI-powered search engine unlocking the collective knowledge of Investigative Reporters and Editors members.
That includes thousands of tip sheets, training materials, conference recordings, slide decks and contest entries.
The site, which launched today at #NICAR26, is powered by semantic search technology. That means it will use AI to match results based on the meaning of your words, not just exact keyword hits.
This is the result of a team effort by a committee of volunteers, with the development led by me, Derek Willis and Cody Winchester.
Access is restricted to active IRE members, authenticated through the recently relaunched ire.org. If you're not a member or let your card lapse, this is a great reason to get out the credit card.
This is an early release and a work in progress. If you encounter bugs, please report them to [email protected]. Here's what we have planned for the month ahead.
For the nerds in the room, the backend is a Qdrant vector database harnessed by Python's FastAPI. Users tap into it via Svelte, the Node.js framework invented at The Guardian.
If you want to talk tech, let's get the comments kicking. We haven't published the code that powers the site, but I'm eager to share the new techniques I learned putting it together.
Aman BhargavaThis is awesome! I'm not a member (yet), but this would have been really cool to use. I also like that you describe what and how much is indexed. Semantic search can feel like a black box sometimes, so this offers some good hints at what is possible. Did you also look into Typesense for search?
Matthew W. Thomas@palewire cool!
I made something recently where I could do access control on the cache. I think it's so much simpler and is so fast. I felt pretty clever for thinking of it. Hope you feel clever too.
Chris Zubak-Skees@palewire Cool, which models?