When:
Wednesday, September 30, 2026 - 15:00 to 15:45 CEST
Room:
Mees Room II
Tags:
development & coding, drupal & open source 101, ai
Track:
SVG
m&b icon_new brand
development, ai & agentic architecture

Scolta: Drop-in AI Search Without a Search Server

Scolta: Drop-in AI Search Without a Search Server

Jeremy Andrews (Jeremy)

What if you could replace Solr with something that runs in the browser, costs almost nothing, and delivers better results out of the box? Scolta is an open-source search toolkit that does exactly that: client-side search powered by Pagefind, with an AI layer for query expansion and summaries. No search server, no vector database, no complex infrastructure. Already deployed on production Drupal sites. Live demo included.

Prerequisite

Familiarity with Drupal's search landscape (core search, Search API, Solr) is helpful but not required. No Rust or WebAssembly experience needed.

Target Audience

Developers, site builders, and technical leads evaluating search solutions for Drupal sites. Also relevant for anyone managing search infrastructure costs or navigating GDPR/data sovereignty requirements.

Outline

Most websites that need decent search end up running something like Solr or Elasticsearch. It works, but it's a lot of infrastructure for "let people find things on my website." And when you want AI features, the conventional path gets heavy fast: RAG pipelines, vector databases, embedding models, and the infrastructure to run all of it.

Scolta takes a different approach. Search runs entirely in the browser using Pagefind, which indexes your content and serves the index as a static asset. No search server. Results in milliseconds. Queries never leave the browser. Scolta then layers AI on top: an LLM handles query expansion, result summarization, and conversational followups.

The architecture is a shared Rust/WASM core compiled to WebAssembly, with a PHP bridge and thin CMS adapters for Drupal, Laravel, and WordPress. One core, three platforms, same behavior.

This session covers:
- How Scolta compares to Solr: what changes, what you drop, where Solr may still fit
- Architecture: why WASM, what the PHP bridge does, how CMS adapters stay thin
- Live demo: browser-side search with zero backend, then AI features turned on
- AI costs: model choice, caching strategies, optimization. Tag1.com runs AI search for under $5/month
- Data sovereignty: search queries stay in the browser (GDPR, HIPAA). AI calls an LLM, so model provider and regional hosting matter
- The Drupal module in depth, plus how the same core serves Laravel and WordPress

Learning Objectives

- Evaluate whether client-side search is a viable replacement for Solr, Elasticsearch or SASS search on their Drupal site
- Understand the Scolta architecture and how a shared Rust/WASM core serves multiple CMS platforms
- Make informed decisions about adding AI search features — cost, data sovereignty, and model selection — to a Drupal site

Experience level
Intermediate