Switch from Algolia DocSearch to Lunr offline search

[vinooganesh]

cc @alamb
Algolia DocSearch loads from cdn.jsdelivr.net which is blocked by Apache's CSP, so search is completely broken on production.

This switches to Docsy's built-in Lunr offline search — fully client-side, no CDN, no API keys. Hugo generates a search index JSON at build time and Lunr searches it in the browser. Follows the Docsy search docs.

Changes

• hugo.toml — enable offlineSearch, remove Algolia config, bump max results to 25 and summary length to 200
• layouts/partials/scripts.html — override removing Algolia JS and markmap CDN blocks

Note: after cloning, run hugo once before hugo server to generate the search index.

Part of #163

[alamb]

Thanks (again) @vinooganesh

I tested this locally and it seems like a step forward to me

ANd it worked great

It is strange that some of the pages don't have any way to click on them 🤔

[vinooganesh]

@alamb fixed! fix was adding a title: to each page

[alamb]

Thanks @vinooganesh

This is looking much better. I merged up from main

One thing I noticed is that there are still some pages that seem to be missing their titles. That being said, this change seems like an improvement to me (and we can perhaps continue to improve things as follow on PRs)

[alamb]

Here is the diff compared to the template

Run this command to compute the diff

diff -du  ~/go/pkg/mod/github.com/google/docsy\@v0.12.0/layouts/_partials/scripts.html layouts/partials/scripts.html

Here is the actual diff:

--- /Users/andrewlamb/go/pkg/mod/github.com/google/docsy@v0.12.0/layouts/_partials/scripts.html	2026-02-11 07:27:43
+++ layouts/partials/scripts.html	2026-02-21 15:07:02
@@ -1,26 +1,18 @@
+{{/*
+  Project-level override of Docsy's layouts/_partials/scripts.html
+
+  Why this file exists:
+  Removes the Algolia DocSearch JS block (cdn.jsdelivr.net) which is blocked
+  by Apache's CSP. Also removes the markmap-autoloader CDN reference since
+  markmap is not enabled on this site.
+
+  See: https://github.com/apache/parquet-site/issues/163
+*/ -}}
 {{ $needKaTeX  := or .Params.math .Site.Params.katex.enable .Params.chem .Site.Params.chem (.Page.Store.Get "hasKaTeX") (.Page.Store.Get "hasmhchem") -}}
 {{ $needmhchem := or .Params.chem .Site.Params.katex.mhchem.enable (.Page.Store.Get "hasmhchem") -}}

-{{ if .Site.Params.markmap.enable -}}
-<style>
-.markmap > svg {
-  width: 100%;
-  height: 300px;
-}
-</style>
-<script>
-window.markmap = {
-  autoLoader: {
-      manual: true,
-      onReady() {
-        const { autoLoader, builtInPlugins } = window.markmap;
-        autoLoader.transformPlugins = builtInPlugins.filter(plugin => plugin.name !== 'prism');
-      },
-  },
-};
-</script>
-<script src="https://cdn.jsdelivr.net/npm/markmap-autoloader"></script>
-{{ end -}}
+{{/* markmap block removed — it loads from cdn.jsdelivr.net which is blocked
+     by Apache's CSP, and markmap is not enabled on this site anyway. */ -}}

 {{ if .Site.Params.plantuml.enable -}}
   <script src='{{ "js/deflate.js" | relURL }}'></script>
@@ -70,25 +62,5 @@
     crossorigin="anonymous"></script>
 {{ end -}}

-{{ if and .Site.Params.search (isset .Site.Params.search "algolia") -}}
-  {{ template "algolia/scripts" .Site.Params.search.algolia -}}
-{{ end -}}
 <script src='{{ "js/tabpane-persist.js" | relURL }}'></script>
 {{ partial "hooks/body-end.html" . -}}
-
-{{ define "algolia/scripts" -}}
-<script src="https://cdn.jsdelivr.net/npm/@docsearch/js@3.8.2"
-  integrity="sha512-lsD+XVzdBI6ZquXc8gqbw0/bgrfIsMJwY/8xvmvbN+U3gZSeG7BXQoCq4zv/yCmntR2GLHtgB+bD4ESPsKIbIA=="
-  crossorigin="anonymous" ></script>
-<script type="text/javascript">
-const containers = ['#docsearch-0', '#docsearch-1'];
-for (let c of containers) {
-  docsearch({
-    container: c,
-    appId: {{ .appId | default "R2IYF7ETH7" }},
-    apiKey: {{ .apiKey | default "599cec31baffa4868cae4e79f180729b" }},
-    indexName: {{ .indexName | default "docsearch" }},
-  });
-}
-</script>
-{{ end -}}

[vinooganesh]

Hey @alamb hmm I'm not able to repro - did you re-run hugo serve after the merge? The index should be recreated upon running that

[alamb]

I thought I did but maybe not. I'll merge this and see how it goes

[alamb]

Thanks again @vinooganesh

[alamb]

Nice -- it is live and appears to be working well

https://parquet.apache.org/docs/