zopyx.typesense 1.0.0a8

Typesense integration with Plone 6

zopyx.typesense

What is zopyx.typesense

zopyx.typesense in an add-on for Plone 6 that provides an integration with the search engine Typesense. The functionality is similar with collective.solr.

The reasons for using Typesense are

• very easy to install (single binary or via Docker)
• multi-field text indexes
• auto-generated (but highly customizable) search UI
• optional facted search for refining fulltext queries
• minimal invasive into Plone (does not replace the ZCatalog)
• very fast search
• extensible and customizable
• scalable (through a Typesense cluster)
• federated search over multiple collections (aka Plone sites)
• open-source
• on-premise or Typesense cloud (commercial offering)

Click here for an intro video.

zopyx.typesense is not

• a replacement of the ZCatalog/portal_catalog (document changes are pushed to Typesense as an independent entity completely decoupled from Plone's portal_catalog)
• a drop-in replacement of the Plone search. zopyx.typesense comes with its own configurable auto-generated search UI

Requirements

• Plone 6 (tested)
• Plone 5.2 (untested, supposed to work for Dexterity-only sites, under Python 3)

Architecture

Typesense runs as dedicated service either on-premise or somewhere in the cloud.

![Architecture](https://github.com/zopyx/zopyx.typesense/raw/master/docs/Typesense.png

Indexing

zopyx.typesense pushes content changes (new documents, updated documents, deleted documents) through Plone lifecycle hooks into Typesense.

Searching

The search interaction between browser and Typesense happens directly without Plone being in-between. When you open the @@typesense-search view, the search configuration (as defined through the search template search.pt and an associated app.js file) is loaded from Plone for auto-generating the search UI (search input, faceted search) within the browser. The search query is directly send to Typesense and the results are directly shown within the browser as soon as they are available. Plone is not involved in the search. No additional filtering (e.g. regarding access control) happens through Plone.

Installation

Add zopyx.typesense to your buildout, re-run buildout and install it within Plone.

For Typesense installation, please check the installation docs of Typesense (either for installation through Docker or through the standalone binary).

There is no public release at this time. You need to install zopyx.typesense using mr.developer as source checkout from Github.

Configuration

The Typesense settings within the Plone controlpanel:

• Name of Typesense collection - must be a unique name for the document pool of your Plone site
• API Key - the administrative API key (as configured in Typesense)
• Search API Key - the search API key (as configured in Typesense)
• URL of Typesense node X - the URL(s) of the Typesense node or Typesense cluster
• Collection schema - the schema of the Typesense collecton (see Typesense docs)

The Typesense administration within the Plone controlpanel:

Example setup

For a typical Plone dev setup which Plone running on localhost, you want to run Typesense (for dev purpose) on the same machine using Docker:

docker run -p 8108:8108 -v /tmp/data:/data typesense/typesense:0.22.1 --data-dir /data --api-key=some_api_key --search-only-api-key=some_search_api_key --enable-cors

This command will store the Typesense data files under /tmp/data and open a connection on port 8108. You need to specify two api keys: one for admin access and one for search api.

After installing zopyx.typesense in Plone, you would specify both API keys in the Typesense control panel within Plone along with http://localhost:8108 as URL for Typesense node 1.

Search UI

The search UI is auto-generated from a minimal HTML template which defines the basic layout together with the widgets to be used see here.

The integration of the template with the Typesense UI generator is impleted through some lines of Javascript code in src/zopyx/typesense/browser/static/app.js.

Views

• @@typesense-search renders the main (auto-generated) search form. This view can be applied on any folder level. This view applied on a folder implies a filter by subpath (only indexed content with the given folder will be searched).

• @@typesense-settings renders the Plone control panel of Typesense

• @@typesense-settings renders the Plone control panel of Typesense

• @@typesense-admin renders the administration control panel for Typesense

• @@typesense-indexed-content applied on any arbitrary content object will display all data indexed for the given content object

Indexing integration

zopyx.typesense integrates into Plone using the lifecycle event for content being added, updated, deleted (plus workflow changes). By default, zopyx.typesense indexes all metadata of all content objects (basically Dublin Core) plus the textual content of a content objects (title, description and the content of all RichText fields). Similar to Plone's SearchableText index, all textual content is being indexed within Typesense in the text field.

It is possible to override the indexing behavior per-content-type using an indexing adapter implementing ITypesenseIndexDataProvider. See here for an example.

Indexing binary content and office formats

The indexer for the File content-type has optional support for indexing binary content or office formats using Apache TIKA. The server URL of the (optional) Tika server must be configured through the Typesense control panel. The most simple approach for running Tika is by using Docker:

docker run -d -p 9998:9998 apache/tika

Configure Tika within the Typesense control panel as http://localhost:9998. Ensure that the configuration regarding host, IP address and port numbers fit your security requirements (usually, you don't want expose an internal service on public IP addresses, but localhost only).

Bulk reindexing

The Typesense admin control panel contains a button Reindex all which will

• create a new collection whose name is defined as <collection_name>-
• submit all Dexterity content objects for reindexing
• alias the new collection with the official name of the configured collection

This means that an existing Typesense collection remains online until the complete reindexing process has finished. The Typesense collection with the fresh index content replaces then the old collection in an atomic operation. There should be no impact on the availability of the search functionality during reindexing.

Custom search views

The default Typesense search view @@typesense-search should work for most common usecases of a Plone site with the default content types. If you need something different, you have two options:

• customize search.pt and app.js according to your own need (e.g. using z3c.jbot)
• create your search view(s) with your own search template and the related JS configuration

Transactions and eventual consistency

All indexing/unindexing operations happen asyncronously to Plone and outside Plone's transaction system. So, content changes might be available in Typesense with a short delay.

Cavecats and known issues

• zopyx.typesense does not integrate (by-design) with Plone's security and access model. The main purpose of zopyx.typesense is to act as a search engine for public sites. So it is recommended at this time to index only public content.

Resources

• Typesense website
• Typesense API
• InstantSearch.js

Author

Andreas Jung | [e-mail] | www.zopyx.com

Paid service for zopyx.typesense is available on request.