PPL Search Intelligence Core

PPL Search Intelligence Core is the self-hosted search backend for Magento Open Source 2.4 and Adobe Commerce 2.4. It is the shared engine of the PPL Search Intelligence system: it builds and serves the search basis and exposes the full admin workflow, but it does not render a storefront by itself. Search Intelligence Core is the backend part and requires exactly one storefront frontend package to show search to shoppers - PPL Search Intelligence for Luma and other classic Magento themes, or PPL Search Intelligence HV for Hyva themes. Both frontends run on this one Core, so the store keeps a single product truth and one admin configuration no matter which theme renders the storefront. It is built for stores where search drives revenue, such as fashion catalogs with color, size, style, and product-type language, and for teams that need explainable, store-owned control over what search returns.

Requires a storefront frontend package (install exactly one)

PPL Search Intelligence Core provides the engine, the admin, and the search AJAX endpoint, but no storefront overlay or search result page on its own. Install one matching frontend for the active storefront surface:

• PPL Search Intelligence - storefront overlay and search result page for Luma and other classic (non-Hyva) Magento themes.
• PPL Search Intelligence HV - storefront overlay and search result page for Hyva themes.

Install exactly one frontend per active storefront search surface. Installed alone, Search Intelligence Core gives you the admin engine, rules, CLI, and the search API, but nothing is drawn on the storefront until a frontend package is installed.

Features (backend engine)

• Product search index on a store-owned MySQL flat basis, with optional OpenSearch product search and Redis/cache-data routing, so the search basis stays on data paths the merchant controls instead of an external search SaaS.
• Catalog-grounded query understanding: product intent, product types, protected words, synonyms, corrections, and vertical knowledge packs, so relevance improves without turning search into an unpredictable suggestion machine.
• Explainable ranking with configurable relevance, freshness, stock, price, product-signal, and optional behavior-signal inputs.
• Zero-result recovery with managed rules and a separate fallback lane for popular products, so a dead-end query keeps the shopper in the buying path.
• Redirect rules for high-intent terms, delivered to the frontend as a suggest-only recommendation (the shopper always chooses to click, so a rule can never trap them).
• Merchandising rules for promoting, demoting, pinning, or shaping product visibility around business priorities.
• Search AJAX endpoint and a single result payload (products, filters, suggestions, recovery, redirect match) that every PPL Search Intelligence frontend renders, so overlay, result page, counts, pagination, and sorting stay aligned to one basis.
• Admin Test Lab for trying real queries and reviewing candidate count, backend mode, timing, payload source, ranking, and recovery before relying on them in the storefront.
• Backend Operations tools for rebuilding product search data, checking search paths, and maintaining runtime readiness, with async indexing support for larger catalogs.
• CLI commands for index rebuilds, knowledge-pack compilation, benchmarks, and diagnostics.
• A multilingual translation baseline and bundled knowledge assets for international search setups.

Why the backend is a separate package

The whole engine (indexing, query understanding, ranking, backend routing, rules, and diagnostics) lives in a single shared Core module and behaves identically whether the storefront runs on Luma or on Hyva. Keeping it in one place, rather than rebuilding it inside each theme, means both frontends share the same product basis, the same admin configuration and rules, the same CLI, and the same single product truth. There is only one backend to build, tune, and maintain, so there is no duplicated development and no drift between the two themes as either one evolves.

It also puts the cost where the value is. The engine, with all of its tuning, lives in Core, which you buy once. The theme frontends are thin adapters that do not carry the engine's cost, so a store that starts on a classic Luma theme and later moves to Hyva already owns Core and only needs to add the slim, low-cost Hyva frontend. If each theme instead bundled its own copy of the engine, both would carry that full price, and changing themes would mean paying for the engine a second time. With one shared Core, your configuration and tuning carry straight over, and you pay for the engine only once.

Why merchants use it

Search tuning becomes a repeatable admin workflow instead of a developer ticket or a guess against a hidden algorithm. Merchants keep one product truth across overlay, result page, filters, pagination, and sorting, so shoppers never see one count in one place and a different count in another. Because the search basis stays on a store-owned MySQL flat index with optional OpenSearch and Redis routing, search behavior stays under the merchant's own control instead of depending on a black-box external search workflow.

Common use cases

• Running one shared search engine behind both a classic-theme storefront and a Hyva storefront, or across a theme migration, without re-tuning search for each theme.
• Fixing typo-heavy and partial-phrase searches with admin-managed synonyms and corrections instead of editing code.
• Pointing high-intent terms to a campaign landing page or category with a redirect rule, then verifying the storefront result before relying on it.
• Diagnosing why a query returns the products it does, using the Test Lab and backend diagnostics during agency handover or support.

What you get

• The installable backend module with the PPL Hub admin area, the search AJAX endpoint, Test Lab, Backend Operations, CLI, configuration, ACL/menu integration, and documentation.
• A PPL Hub admin area covering general behavior, suggestions, search page, backend routing, data routing, index behavior, Redis, OpenSearch, related terms, smart discovery, merchandising, ranking, signals, and storefront UI display options that the frontend consumes.
• Admin Test Lab and Backend Operations for query testing, rebuilds, backend checks, and index readiness.
• Synonym, correction, zero-result, redirect, and merchandising rule management.
• MySQL flat index support, optional OpenSearch 2.x / OpenSearch 3.x product search, Redis/cache-data routing, and explicit fallback controls.
• A multilingual translation baseline and bundled knowledge assets for international search setups.
• Installation Guide, User Guide, and Reference Manual PDF documentation.
• Note: the storefront search overlay and search result page are delivered by the frontend package (PPL Search Intelligence or PPL Search Intelligence HV), not by Search Intelligence Core.

Compatibility

Designed for Magento Open Source and Adobe Commerce 2.4 environments, with PHP and package requirements declared in composer.json. Optional OpenSearch runtime paths support OpenSearch 2.x and OpenSearch 3.x when the installed 2.4 release supports that service version. PPL Search Intelligence Core uses PPL Hub for shared admin navigation. It requires exactly one storefront frontend - PPL Search Intelligence (classic themes) or PPL Search Intelligence HV (Hyva themes) - to produce storefront output; do not install both frontends for the same active storefront search surface. PPL Discovery Filters can be used as an optional compatible filter partner, and aggregated behavior signals from PPL Tracking are consumed only when that module is installed and allowed by policy. PPL Scored Search is an alternative search engine and must not be enabled together with the Search Intelligence stack for the same storefront search surface.

Ideal for

Fashion and language-rich catalogs, and the merchants and agencies running them, who want one store-owned search engine shared across Luma and Hyva storefronts - with explainable ranking, admin-managed merchandising, and a single product truth - and who add the matching PPL Search Intelligence frontend for the theme they run.