Eric Pugh

YouTube video by Eric Pugh Quepid Working Draft Demo

The best intro to Quepid is still this one from TEN years ago: www.youtube.com/watch?v=5KdN...

OpenSearch's Learning to Rank (LTR) plugin had a busy year! @sstults.bsky.social shares a year in review: opensourceconnections.com/blog/2025/12.... #opensearch #learning_to_rank

Now playing with Ollama and qwenn 30b locally. Much faster then my old laptop.

My favorite laptop was the 12" Macbook Pro, and the Air is the closest I can get.

First new mac in six years, Macbook Air M4 with 32gb of ram. Compared to my old intel one, Solr testsuite in 18 minutes instead of 52 minutes. "tidy" in 18 seconds compared to 52 seconds.

OpenSearch UI means... working on my React skills with my buddy Claude... I never thought I would add "React" to my list.... Sigh.

AI works best in processes where evaluation and trial and error is already in the process today. That’s why it fits into coding so well.

From the Charlottesville community on Reddit Explore this post and more from the Charlottesville community

The BEST conference in Virginia (other than Haystack of course): www.reddit.com/r/Charlottes...

Argh. #opensearchcon

Nice visual of AI Agent concept in Deep Research at the #openseachcon Unconference. Thanks Jakub of @zetaalpha for sharing.

Do you collect feature adoption or other more user centric data? Data that feeds back into how you evolve the product? I think I may be asking “how do you collect data that informs your product roadmap”

What we found however at our clients is that Analytics teams don’t speak “Oberservability” and use a different tech stack.

Last year we did a spike to funnel user click data like add to cart or click through s via OpenTelemetry and leveraging oberservability features of OpenSearch and it was promising.

Has the observabikity world ever moved up the “Business Stack” to proactively monitor business events? Seems like all that infrastructure for monitoring systems is almost ther to measure your BusinessOps? #opensearch #opensearchcon. Tell me what exists out there right now!

Want more Python and less JSON for building algorithms for search in #OpenSearch? Buttonhole these members of the Tech Steering Committee! #opensearchcon

I was the original host of the Hackathon, and then a workshop for work reared it's ugly (actually, totally exciting) head and I had to rely on my community for help! And the community came through.

Thanks David Smiley and Jason Gerlowski for hosting the Search Track Hackathon at @apache.org Community/Code conference:

SIGIR LiveRAG Challenge Report - OpenSource Connections OpenSource Connections participated in the SIGIR 2025 LiveRAG Challenge, where teams built RAG systems in 2 hours to answer 500 questions.

My colleagues did the SIGIR challenge…. opensourceconnections.com/blog/2025/08...

GitHub - seuros/rails_lens: Comprehensive Rails application visualization and annotation tool - precision optics for the Rails universe Comprehensive Rails application visualization and annotation tool - precision optics for the Rails universe - seuros/rails_lens

Just saw a tool to explain hour Rails app to a LLM. I love the humor in the readme ;-). github.com/seuros/rails...

Add Import Capability for externally run Search Evaluation by epugh · Pull Request #120 · opensearch-project/search-relevance Description This addresses Issue #46 which allows importing of documents and evaluation results. This is a re-implementation of the work that @sstults did in #73 after the discussion about using th...

This is an amazing Code Review comment! I didn't appreciate how good it is until re-reading it with @sstults.bsky.social. Thanks Martin! I aspire to this level.

github.com/opensearch-p...

Also, check out Superlinked as a option that drives towards “it doesn’t matter so much the vector db as in how you use” example.

Vector DB Comparison Vector DB Comparison is a free and open source tool from VectorHub to compare vector databases.

In some ways I think there are many reasonable choices for the BUILD option…. superlinked.com/vector-db-co... is a good resource. Maybe more important than which db is HOW you use it.

Ugh, I just asked an LLM to count the number of words in a article I am writing (goal is 2000!). 10 to 15 seconds later, after it read every word, it did give me an answer... Or I could have done "wc -w fast_accurate_relevant_intuitive.md" ;-)

Are you asking ones that DO the extraction or ones that you can use to BUILD your own extractions with?

Using Claude on Rails to write large swaths of code, but Zed's built in Claude integration to tweak/fine tune code on a more file by file basis...

Interested in Team Draft Interleaving for A/B testing? Check out this RFC by @davidfisher1960.bsky.social github.com/opensearch-p...

GitHub - a-chacon/oas_rails: Generate Automatic Interactive Documentation for Your Rails API Generate Automatic Interactive Documentation for Your Rails API - a-chacon/oas_rails

Quepid now has lovely API docs thanks to github.com/a-chacon/oas..., my favorite OpenAPI compliant documentation tool for Rails!

I love that at @mices.co BarCamp that each group does a quick read out to everyone so I do lt get FOMO!

My highschol science fair project was a HyperCard based simulation of dissecting a fetal pig…. Never could have imagined the project without what Bill Atkinson invented. Thank you. m.slashdot.org/story/442967

Hyperwood Objects in three-dimensional space made of two-dimensional slats, generated by parameterized algorithms.

Time for another interesting format: The Hyperwood Exchange Format (HEF).

Based on the cubicle exchange format, coming from an 3d editor based on voxels (remember Comanche?), HEF is a language for building furnitures. And Hyperwood is the idea of open source furniture.