2026-01-06
As we stepped into 2026, we are changing gears to also improve your experience on Boiling Steam. This time, we are launching a search engine on our platform so that you can more easily retrieve previous articles we have published since 2014.
Access the search engine here:
https://boilingsteam.com/search/
As we are now running a static website, you may be wondering how this feature actually works. It’s not that complicated in terms of structure:
- Every time a new article is published, an index using the BM25 algorithm is re-created from scratch. It exists as a flat JSON structure.
- That JSON structure is compressed to take the smallest footprint possible. It ends up taking less than 2 Mb at the time of writing.
- On the [search p…
2026-01-06
As we stepped into 2026, we are changing gears to also improve your experience on Boiling Steam. This time, we are launching a search engine on our platform so that you can more easily retrieve previous articles we have published since 2014.
Access the search engine here:
https://boilingsteam.com/search/
As we are now running a static website, you may be wondering how this feature actually works. It’s not that complicated in terms of structure:
- Every time a new article is published, an index using the BM25 algorithm is re-created from scratch. It exists as a flat JSON structure.
- That JSON structure is compressed to take the smallest footprint possible. It ends up taking less than 2 Mb at the time of writing.
- On the search page, provided you run Javascript, on the client side, the page loads the compressed index and uncompresses it on the fly, and can match the keywords you enter as part of your search query with the contents of the index with the BM25 algorithm. Results are displayed in real-time as you type.
To make things a little nicer and more relevant, when there are matches directly in the titles of articles, there is a higher weight given to the results. Optionally, through a checkbox, you can decide to give more weight to recent articles to re-rank the results.
This is by no means a perfect method, but it’s robust enough for a great number of searches. And in our case, it’s better than the non-existent search capability that was the statu quo until now!