README

MageSetu Semantic Search is a Magento 2 module that adds AI-powered semantic product discovery alongside Magento’s native catalog search. It indexes product content, generates vector embeddings asynchronously, and retrieves relevant products from vector stores such as OpenSearch or Qdrant.

Why this module

This module is designed for merchants and developers who want an intelligent search experience that goes beyond keyword matching. It helps surface products based on meaning, context, and similarity rather than exact text matches.

Key features

Semantic product indexing for catalog content
Asynchronous embedding generation via Magento message queues
Support for OpenSearch and Qdrant vector storage backends
Admin configuration for embeddings, guardrails, personalization, and vector settings
Frontend AI-style search experience with a dedicated storefront flow
CLI maintenance commands for orphan cleanup and environment prefix inspection
Detailed module documentation for configuration, services, and architecture

Requirements

PHP 8.4 or newer
Magento 2.4.8 or newer
Composer dependencies:
- magesetu/module-common
- magesetu/module-ai-base

Installation

Composer-based installation

composer require magesetu/module-semantic-search
bin/magento module:enable MageSetu_SemanticSearch
bin/magento setup:upgrade
bin/magento cache:flush

Local module development

If you are working from this repository directly, place the module under your Magento installation in the expected module path and run the same setup commands.

Configuration

After installation, configure the module from the Magento admin panel:

Enable the module
Select the embedding provider and model
Choose the vector storage backend
Configure guardrails and personalized-message options
Set a collection prefix for multi-environment isolation if needed

The detailed configuration guide is available in docs/configurations/adminhtml/system-config-and-defaults.md.

Usage

Once enabled, the module will:

Collect semantic product content during indexing
Prepare payloads for embedding generation
Send work to the queue for asynchronous processing
Store embeddings in the selected vector database
Serve semantic results through the AI search experience

You can also use the CLI commands documented in docs/features/semantic-search/console-commands.md to inspect or clean up vector data.

Architecture at a glance

The module is organized around a few core stages:

Indexer flow: prepares product data for semantic processing
Payload processing: decides whether content needs to be re-queued
Embedding generation: converts product context into vectors
Vector storage: stores and retrieves similar product embeddings
Frontend search flow: handles the storefront request and response experience

Documentation

The repository includes a structured documentation set under docs covering configuration, services, features, and database schema.

Recommended starting points:

docs/index.md — documentation hub
docs/features/semantic-search/frontend-search-experience.md — storefront flow
docs/features/semantic-search/indexer-flow.md — indexing pipeline
docs/features/semantic-search/console-commands.md — maintenance and diagnostics
docs/configurations/adminhtml/system-config-and-defaults.md — admin configuration reference

Disclaimer

Warning

Use this module at your own risk. This module is provided "as is" without warranty of any kind, either express or implied. The developers are not responsible for any data loss, API rate limit overages, financial charges from external LLM providers (e.g. OpenAI), server downtime, or other issues resulting from the use or installation of this software.

License

This project is licensed under the Apache License 2.0. See LICENSE for details.

magesetu / module-semantic-search

Package info

Security

Statistics