Czy roboty AI nie poradzą sobie bez tego?

Kowal AI Product Feed to moduł Magento 2, który zamienia dane produktowe sklepu w uporządkowaną, aktualną i gotową do wykorzystania bazę wiedzy dla systemów AI. Jego zadaniem nie jest zwykły eksport katalogu, ale przygotowanie treści w taki sposób, aby mogły być bezpiecznie i skutecznie używane przez rozwiązania oparte o OpenAI Vector Store, wyszukiwarki semantyczne, asystentów AI oraz mechanizmy RAG. Jeśli potrzebujesz bardziej szczegółowej odpowiedzi, możesz doprecyzować pytanie lub skorzystać z klasycznego formularza kontaktowego.

Skip to the end of the images gallery

Skip to the beginning of the images gallery

Kowal AI Product Feed for OpenAI Vector Store and LLMs

Be the first to review this product

€61.50 €50.00

COMPOSER Installation

M2-AI-FEED-VECTOR-STORE

Requires changes to the template

Minor changes

Significant changes

Requires programming knowledge

Basic

Advanced

Difficulty in configuration

Impact on performance

Compliance with Magento standards

Details

Kowal AI Product Feed is a Magento 2 module for building an AI data layer that prepares store content in a format ready to be used by OpenAI Vector Store, AI assistants, semantic search engines, RAG pipelines, and public content discovery mechanisms such as llms.txt and NDJSON feeds.

This is not a standard product catalog export. The module organizes, normalizes, and publishes store knowledge in a way that allows it to safely power AI systems while maintaining control over data freshness, content scope, and document structure.

As a result, Magento can become a real source of knowledge for:

product chatbots and copilots,
intelligent customer service,
semantic search,
AI-generated answers,
content automation,
external AI integrations based on public or private feeds.

What the module does in practice

The module retrieves data from Magento and builds structured knowledge documents from it. It supports not only product content, but also additional content channels that are important from both AI and sales perspectives.

It can currently work with content such as:

product.core - core product data, descriptions, URL, attributes, and categories,
product.faq - product questions and answers,
product.docs - documentation and Markdown files,
blog.article - blog posts,
category.content - category content,
cms.page - CMS pages.

Each document is saved locally, compared by checksum, and synchronized only when it has actually changed. This reduces redundant uploads, stabilizes the synchronization process, and provides greater control over costs and data quality.

Public AI and LLM endpoints

In the new version, the module can also publish data under public frontend endpoints:

/llms.txt
/ai-feed/{storeCode}/index.json
/ai-feed/{storeCode}/{contentType}.ndjson

This means the store can not only synchronize data to OpenAI Vector Store, but also provide a controlled content layer for external AI agents, crawlers, and integration tools.

The module can additionally:

add a signal to the page head through a link to llms.txt,
optionally display a link to llms.txt in the footer,
publish feeds per store view and per language,
maintain consistency between store content, the AI layer, and the public content discovery layer.

Why this matters for business

In many stores, the data needed to implement AI is scattered: some is in product descriptions, some in FAQ, some in the blog, and some in separate technical materials. Without an integration layer, it is difficult to build stable AI solutions on top of that.

Kowal AI Product Feed solves this problem because it:

centralizes store knowledge in a single export and synchronization architecture,
organizes data for use by language models,
reduces the cost of building additional AI integrations,
shortens the implementation time for chats, assistants, and semantic search engines,
makes it possible to develop an AI SEO, AEO, and LLM discoverability strategy,
provides a foundation for multilingual AI implementations in Magento 2.

This means faster rollout of new features, greater consistency of AI-generated answers, and better use of the content the store already has.

Who this module is for

The module is designed for:

Magento 2 stores implementing AI chat or an AI assistant,
companies building semantic search engines and RAG on e-commerce data,
technical teams integrating Magento with OpenAI,
stores developing multilingual product knowledge bases,
agencies and software houses implementing AI in Magento,
brands that want to prepare their store for future LLM content discovery channels.

How other modules use it

Other modules can use Kowal AI Product Feed as a central knowledge source and integration layer instead of building separate exports, custom data serialization, and their own synchronization processes.

Example use cases:

an AI chat module can ask questions to the Vector Store and answer based on current product knowledge,
an FAQ module can enrich the shared knowledge base with additional questions and answers,
a documentation module can feed AI with instructional and technical content,
a customer service module can use the same data for contextual answers,
a recommendation module can use structured content for better product matching,
content modules can extend the knowledge layer with blog, CMS, and additional document types.

Most importantly, other modules do not need to know the technical details of synchronization, checksums, file export, or upload to OpenAI. They receive a ready-to-use, consistent, and extensible data layer.

Key benefits

a ready AI data layer for Magento 2,
support for OpenAI Vector Store,
public llms.txt and NDJSON feeds,
synchronization of changed content only,
support for multiple content channels,
prepared for AI SEO, AEO, and LLM integrations,
compatible with Magento store view architecture and language versions,
a solid foundation for further AI automation development in the store.

Module Installation Instructions

Kowal AI Product Feed - installation and configuration

Document purpose

This document describes the complete process of installing, launching, and configuring the Kowal_AiProductFeed module in Magento 2. It has been prepared so you can go from technical implementation to the first working export and the publication of llms.txt.

Module functional scope

The module enables:

exporting store content to JSON and JSONL files,
synchronizing data to OpenAI Vector Store,
support for multiple content channels such as product, blog, category, and CMS,
publishing public endpoints:
- /llms.txt
- /ai-feed/{storeCode}/index.json
- /ai-feed/{storeCode}/{contentType}.ndjson
adding a link to llms.txt in the head section,
optionally adding a link to llms.txt in the store footer.

llms.txt works per current Store View. This means each language version can return its own manifest and its own feed links.

Prerequisites

Before installation, make sure that:

Magento 2 is working correctly,
you have SSH access to the Magento root directory,
Composer is available in the environment,
the PHP process can write to the var/ directory,
you have an active OpenAI API Key if you want to use synchronization to OpenAI Vector Store.

Module installation

1. Adding the Composer repository

If the module is installed from a Git repository:

composer config repositories.ai.product.feed vcs https://github.com/kowalco/module-ai-product-feed

If the repository is private:

composer config --global --auth github-oauth.github.com

2. Package installation

composer require kowal/module-ai-product-feed

3. Enabling the module

bin/magento module:enable Kowal_AiProductFeed

4. Updating Magento

bin/magento setup:upgrade

5. Compilation and cache

In a production environment, also run:

bin/magento setup:di:compilebin/magento cache:flush

In a development environment, it is usually enough to run:

bin/magento cache:flush

If there is an additional cache layer in front of Magento, also perform a purge on the side of:

Varnish
Cloudflare
Nginx fastcgi/proxy cache
another reverse proxy or CDN

6. Confirming module status

bin/magento module:status Kowal_AiProductFeed

Configuration location

The module configuration is available in the admin panel:

Stores -> Configuration -> General -> Kowal AI Feed

Recommended configuration model

The best practice is to configure at the Store View level, especially if the store operates in multiple languages.

Recommendation:

one Store View = one language,
one Store View = one Vector Store ID,
one Store View = a separate public AI feed layer.

Example:

PL store view -> Polish Vector Store ID
EN store view -> English Vector Store ID

Field configuration

`Enable Module`

Enables the module for the selected scope.

Recommendation:

set Yes for those Store View entries that should generate AI data

`OpenAI API Key`

Global API key used to communicate with OpenAI.

Required if:

you want to synchronize files to OpenAI Vector Store

Not required if:

you only want to use local exports,
you only want to use llms.txt and public feeds.

`Vector Store ID`

The OpenAI Vector Store identifier for a specific Store View.

Recommendation:

configure at the Store View level,
set a separate identifier for each language.

`Feed Location`

Relative path to the directory where export files will be saved.

Default value:

var/ai-feeds

Recommendation:

keep the directory in var/,
do not expose this directory directly through the web server.

`Enable Public AI Feed`

Enables public endpoints:

/llms.txt
/ai-feed/{storeCode}/index.json
/ai-feed/{storeCode}/{contentType}.ndjson

Enable this option only for those store views whose content may be publicly available.

Important:

/llms.txt is not a global index of all stores
each language version of the store returns its own manifest
/llms.txt itself does not require generated files to exist in order to return a valid response

`Show LLMs.txt Link in Footer`

Adds a visible link to /llms.txt in the store footer.

Recommendation:

No by default,
Yes if you want to explicitly communicate the availability of the AI layer to partners, integrators, or bots.

`Docs Location`

Relative path to the directory with Markdown product documentation for product.docs.

Examples:

var/ai-docs
var/ai-docs/pl

Supported files include:

{docsLocation}/{SKU}.md
{docsLocation}/{SKU}.markdown
{docsLocation}/{storeCode}/{SKU}.md
{docsLocation}/{storeCode}/{SKU}.markdown
{docsLocation}/{SKU}/*.md
{docsLocation}/{SKU}/*.markdown
{docsLocation}/{storeCode}/{SKU}/*.md
{docsLocation}/{storeCode}/{SKU}/*.markdown

`Max Products Per Batch`

Batch size for product export.

Starting value:

500

Reduce the value if:

the catalog is very large,
the environment has limited memory,
additional AI operations increase the load.

`Sync Product Core`

Enables the product.core content type.

Recommendation:

Yes

`Sync Product FAQ`

Enables product.faq.

Use if the store has data in the kowal_zapytajoprodukt_pytania table.

`Sync Product Docs`

Enables product.docs.

Use if the Markdown documentation is already prepared.

`Sync Blog Content`

Enables the blog channel and the blog.article type.

`Sync Category Content`

Enables the category channel and the category.content type.

`Sync CMS Pages`

Enables the cms_page channel and the cms.page type.

`Max Retries`

Maximum number of API operation attempts.

Default:

`Polling Timeout`

Maximum wait time for file processing to finish on the OpenAI side.

Default:

600

`Polling Interval`

Interval between subsequent status checks.

Default:

`HTTP Timeout`

Maximum duration of a single HTTP request to OpenAI.

Default:

60

Recommended configuration order

After installation, configure the module in this order:

Enable the module for the target Store View.
Add the OpenAI API Key if you plan to synchronize with OpenAI.
Set the Vector Store ID for each language.
Verify the Feed Location.
Set the Docs Location if you use Markdown documentation.
Enable the required content types and channels.
Enable Enable Public AI Feed if the endpoints should be publicly available.
Optionally enable Show LLMs.txt Link in Footer.
Clear the Magento cache.

bin/magento cache:flush

First launch

Dry run

First perform a dry run for one Store View:

bin/magento kowal:ai-export:sync --channel=product --store=1 --dry-run

Actual export

Then run the actual synchronization:

bin/magento kowal:ai-export:sync --channel=product --store=1

Status check

bin/magento kowal:ai-export:status --store=1

Verification of public endpoints

If Enable Public AI Feed is enabled, check:

curl -i https://twoja-domena.pl/llms.txtcurl -i https://twoja-domena.pl/ai-feed/pl/index.jsoncurl -i https://twoja-domena.pl/ai-feed/pl/product.core.ndjson

If the store has separate language versions in the URL, also check:

curl -i https://twoja-domena.pl/en/llms.txtcurl -i https://twoja-domena.pl/de/llms.txt

Expected behavior:

/llms.txt returns a manifest only for the current Store View
/llms.txt works even when feed files have not yet been generated
/ai-feed/{storeCode}/index.json shows only existing feeds for the given store
/ai-feed/{storeCode}/{contentType}.ndjson works only when the relevant source files exist

Cache header verification:

curl -I https://twoja-domena.pl/llms.txtcurl -I https://twoja-domena.pl/ai-feed/pl/index.json

You should see:

for /llms.txt, headers that practically disable cache,
for index.json, a short cache lifetime and forced revalidation.

HTML verification:

the page source should contain a link in the head to /llms.txt,
if you activated the footer option, an LLMs.txt link should appear in the footer.

CLI commands

Synchronization

bin/magento kowal:ai-export:sync --channel=productbin/magento kowal:ai-export:sync --channel=product --store=1bin/magento kowal:ai-export:sync --channel=product --store=1 --content-type=product.core --identifier=example-sku --dry-runbin/magento kowal:ai-export:sync --channel=blog --store=1 --source=amasty_blogbin/magento kowal:ai-export:sync --channel=category --store=1bin/magento kowal:ai-export:sync --channel=cms_page --store=1

Status

bin/magento kowal:ai-export:statusbin/magento kowal:ai-export:status --store=1bin/magento kowal:ai-export:status --channel=blog --store=1bin/magento kowal:ai-export:status --channel=blog --store=1 --source=amasty_blog

Import blog sources

bin/magento kowal:ai-blog-source:import --source=amasty_blog --store=1

Legacy product alias

Still available for backward compatibility:

bin/magento kowal:ai-feed:generatebin/magento kowal:ai-feed:generate --store=1 --content-type=product.core --sku=example-sku --dry-run

Cron

The module uses three cron jobs:

kowal_ai_blog_source_import
kowal_ai_feed_generate
kowal_ai_feed_cleanup

Make sure Magento cron is working correctly in the environment.

Most common issues

`/llms.txt` returns `404`

Check:

whether the current module version with frontend routing is deployed,
whether you ran bin/magento setup:upgrade,
whether in production mode you ran bin/magento setup:di:compile,
whether you performed the standard static content deployment according to the project process,
whether you cleared the cache,
whether you purged the CDN / reverse proxy layer,
whether Enable Public AI Feed is enabled.

If https://twoja-domena.pl/llms.txt?v=123 works but https://twoja-domena.pl/llms.txt does not, this usually means an old shared cache is still active somewhere in the chain.

`/llms.txt` works, but the feeds are empty

This usually means that:

the export has not been run yet,
there is no active content for the given channel,
the relevant synchronization flags are disabled.

This behavior is correct. The manifest itself can work before the actual feeds do.

No files in `var/ai-feeds`

Check:

whether the module is active for the correct Store View,
whether the required content types are enabled,
whether Magento has write permissions to var/,
whether the synchronization process has been started.

Problems with synchronization to OpenAI

Check:

the correctness of the OpenAI API Key,
the correctness of the Vector Store ID,
the availability of outgoing HTTP connections,
timeout and retry values.

Summary

After proper installation and configuration, the module provides Magento 2 with a ready AI data layer:

local,
multilingual,
capable of synchronization with OpenAI,
ready for publishing through llms.txt and NDJSON feeds,
extensible for additional integrations and AI modules.

Change Log

Changelog: Initial release includes product data normalization, per-product document generation, JSON and JSONL export, OpenAI Vector Store synchronization, checksum-based change detection, local sync registry, and cleanup of obsolete remote files.

Questions and Answers

Question

Can't AI robots manage without this?

Answer

Kowal AI Product Feed is a Magento 2 module that turns a store's product data into a structured, up-to-date, and ready-to-use knowledge base for AI systems. Its purpose is not a standard catalog export, but to prepare content in a way that allows it to be safely and effectively used by solutions based on OpenAI Vector Store, semantic search engines, AI assistants, and RAG mechanisms. If you need a more detailed answer, you can clarify your question or use the standard contact form.