Introduction
For many years, searching and browsing within the storefront was done with Solr. Solr will be replaced by SPARQUE.AI.
This concept describes the MVP of the integration into the Intershop ecosystem. Hint: it is very different from the Solr integration.
The full integration will be built in agile fashion. This document describes the MVP of the integration - which in the future will further grow, and also might change in some aspects.
Glossary
SPARQUE, SPARQUE.AI | AI-powered search solution provided by Intershop |
|---|---|
SPARQUE workspace | Encapsuled environment within SPARQUE |
PWA | Progressive Web App |
Graph, Knowledge Graph | SPARQUE term for search index. Contents from the data sets are combined into a structure which allows for making very efficient search queries. The structure of the graph can be defined very freely within the SPARQUE desk. |
References
Guide - Setup a Project with SPARQUE
Architecture Overview
Involved Components
Component | Notes |
PWA | Usually all REST requests from the PWA go to the ICM. With the SPARQUE integration, some requests now are routed to SPARQUE. |
Policy Enforcer | Service between PWA and SPARQUE. Required to ensure requests (esp. personalized request) are properly authenticated. |
SPARQUE | AI-powered search solution provided by Intershop. |
Data Provider | SPARQUE needs always complete data sets. For this purpose, the Data Provider stores the most recent version of all files in one location. From this place SPARQUE can pull the data easily and build a new graph. For the MVP we use Azure file storage. |
ICM | Creates JWT Token including User ID and Channel. With that additional information the Policy Enforcer can authorize the search requests. |
Versions and Compatibility
PWA | ICM | SPARQUE Policy Enforcer Data Provider |
|---|---|---|
https://github.com/intershop/intershop-pwa/tree/feature/sparque feature branch can be used in project | ICM 11.2.0 | will be provided as services from Intershop |
Features
PWA capabilities (via API calls)
browsing category pages (i.e., pages with sub-categories)
product search suggest
product search (without filtering)
One search graph per channel
Full graph build
scheduled
manually started
REST API
Some REST calls go directly to SPARQUE to retrieve information. They are not routed through ICM.
All other requests (customer data, baskets, cms, etc.) continue to be handled by ICM.
The standard integration provides a number of endpoints which fit perfectly into the PWA.
Known Limitations
Currently the Intershop-Sparque integration does not have full feature parity with the existing Solr-based functionality. Since the concepts of Solr and SPARQUE are different, this state will not be reached - to keep the flexibility of SPARQUE, there will always remain customization for each project.
This integration is intended as prefabrication - there is still some work to be done in the project. Over time, the prefab part will grow.
The following features are not supported at the moment. Some of them will be added in the future.
Limitation | Details | Hints for project development | |
|---|---|---|---|
| 1 | ICM 7.10 support | The authentication via JWT SPARQUE extension was implemented for ICM 11. Can be added to 7.10 in a project. |
|
| 2 | Responsive Starter Store support | The integration was built into the PWA only. RSS support is not on the roadmap. |
|
| 3 | Configuration (filters and facets, synonyms, etc.) | Since the integration is mostly outside the ICM, the ICM back office cannot be used for configuration. A new GUI is needed, and will be built in either SPARQUE desk or the Intershop Administration Portal. On the roadmap. |
|
| 4 | 10+ Channels | We plan to put data from 1 channel into 1 data set, to make sure that ICM product sharing is fully supported. SPARQUE has no experience of this kind of setup for a high number of channels. Possibly this setup does not scale well for graph rebuilds and API updates. |
|
| 5 | Personalized searching and browsing | For the MVP, we only use product and catalog data. For personalization, additional data is required, e.g. order or tracking data. Can be added in project. |
|
| 6 | Partial graph updates | Limitation within SPARQUE. The way the SPARQUE knowledge graph is calculated basically requires a full rebuild. No partial or real-time updates are possible. |
|
| 7 | Full end-to-end process automation | Automation will grow over time. On the roadmap. |
|
| 8 | Compatibility with ICM staging | On the roadmap. |
|
| 9 | Storefront features | The current feature set is limited. Some missing elements are:
On the roadmap. |
Setup and Configuration
As described above, the integration connects many separate elements. These are lying in various areas of responsibility. Some need to be setup individually, others just need additional configuration.
Follow Guide - Setup a Project with SPARQUE for details.
Search Graph Building Process
SPARQUE works very differently from Solr. We recommend the lecture at SPARQUE Desk Documentation and/or user training, to familiarize yourself with the concepts.
Customization Options
SPARQUE has powerful configuration options. Refer to Guide - Setup a Project with SPARQUE for an overview of possibilities that can be used in projects.