<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title>Pgvector - vo.rs</title><link>https://vo.rs/tags/pgvector/</link><description>Latest from the Pgvector desk at vo.rs.</description><generator>Hugo -- gohugo.io</generator><language>en</language><copyright>This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.</copyright><lastBuildDate>Mon, 20 Jan 2025 11:33:00 +0000</lastBuildDate><atom:link href="https://vo.rs/tags/pgvector/" rel="self" type="application/rss+xml"/><item><title>Running Embeddings Locally for Self-Hosted Search</title><link>https://vo.rs/story/running-embeddings-locally-for-self-hosted-search/</link><description>&lt;p&gt;Keyword search fails in a very specific and irritating way. I search my notes for &amp;ldquo;server won&amp;rsquo;t boot&amp;rdquo; and find nothing, because the note I wrote three years ago says &amp;ldquo;the box refuses to POST after a power cut&amp;rdquo;. Same problem, no shared words, and full-text search — which matches strings — has no idea the two are related. Every self-hosted search box I&amp;rsquo;ve run has this hole: it finds documents that contain your words and misses documents that contain your meaning. Embeddings are how you close it, and the model that produces them will run happily on the same homelab hardware as everything else.&lt;/p&gt;</description><pubDate>Mon, 20 Jan 2025 11:33:00 +0000</pubDate></item></channel></rss>