AI Product Recommender and Bundles for WooCommerce – Related Products, Complete the Look & Add All to Cart
Related products and “complete the look” bundles ranked by AI meaning similarity and what your customers bought together, with Add all to cart, a preview with the reasons, pins and exclusions.
AI Product Recommender and Bundles ranks related products and “complete the look” bundles with two signals: the meaning of your product texts (AI embeddings) and what your customers bought together (counted on your site from your own orders), blended with weights you choose.
On the shop
Related products on the product page: replace the WooCommerce related products (your theme's look is kept) or add a section
A “Complete the look” bundle with checkboxes, the total and an Add all to cart button
Suggestions on the cart page and, optionally, the order received page
A block and a shortcode
Add all to cart, done carefully
Every product is checked again on the server: part of the bundle of the page, visible, purchasable, in stock. Variable products need a chosen option (otherwise they are skipped and the shopper is told), quantities are capped, and prices always come from WooCommerce. There is no bundle discount in this plugin: use your pricing plugin if you want one.
Know why
The admin Preview shows, for any product, what visitors would get with the similarity %, bought together N times, the pinned and fallback products, and the products left out with the reason.
Your rules
Pins per product (shown first) and Never recommend
Exclude categories, tags, products and out-of-stock products; hidden, private, draft and password-protected products are never shown
Fallback to related products by category and tag when there is no AI or order data
Data and costs
Embeddings in background batches with progress and an estimate; a product is sent again only when its text changes
The similar products are computed on your server; a page view never calls the AI and the ranked lists are cached and cleared on change
Order counts are computed locally (HPOS and legacy storage): only product pairs and counts are stored, no customer data is stored or sent
No tracking of visitors and no cookies of the plugin; a daily AI limit (1,000 by default)
Optional explanation line
Off by default: one short line “why these go together” per bundle, written once in the background, kept, and shown only after a person approved it.
Honest limits
Measured on a test machine: lists for 1,000 products in about 20 seconds to 1.6 minutes of background work, for 10,000 products about 29 minutes (vector size 256) to 4.5 hours (1536). Above that a vector database is the better tool. Suggestions are not personalised. The AI can be wrong: check important products in the preview. The screenshots were taken on a demo shop with demo orders and a deterministic test embedding service (not a real model).
Tested
151 automated checks: ranking with deterministic test embeddings and synthetic orders, blending, exclusions, pins, Add all to cart validation and hostile input, HPOS and legacy order storage, cache invalidation, atomic daily limit, permissions, uninstall.
Documented
Illustrated guide in English and Arabic, and an Arabic translation of the plugin.
What's included
ai-product-recommender-and-bundles-for-woocommerce.zip: the WordPress plugin (translation-ready, Arabic translation included)
Illustrated documentation in English and Arabic (HTML and PDF)
Licence file (GPL-2.0-or-later) and changelog
Requirements
WordPress 6.0+, PHP 7.4+, WooCommerce 8.0+ and, for the meaning similarity, an API key of an OpenAI-compatible provider with an embedding model.
What's in the download
The files in the archive you can download after you buy.
No reviews yet. Bought this item? Be the first to review it.
Usage
Install and use it in a few lines.
// The recommendations from your own code (ids, scores, reasons)$res = TQPR_Engine::recommend( $product_id, 'bundle', 3 );// Add a brand or any meta to the text that is embeddedadd_filter( 'tqpr_product_text', function ( $text, $product ) { return $text . "\nBrand: " . $product->get_meta( 'brand' );}, 10, 2 );// Never recommend some products (return a short reason)add_filter( 'tqpr_visible', function ( $ok, $product ) { return $product->get_meta( 'b2b_only' ) ? 'b2b' : $ok;}, 10, 2 );// Highest quantity of one product added by the bundle button (default 10)add_filter( 'tqpr_bundle_max_qty', function () { return 3; } );// The key can live in wp-config.phpdefine( 'TQPR_AI_KEY', 'sk-...' );
Documentation
Setup, settings and troubleshooting, written by the author.
Related products and “complete the look” bundles from AI embeddings blended with bought-together counts
Bought-together counts computed locally from completed orders (HPOS and legacy storage), counts only
Add all to cart with server-side validation, variations, stock and capped quantities
Admin preview with the reasons, pins, exclusion rules, fallback by category and tag
Background indexing with progress, re-index on save, cache with invalidation, atomic daily AI limit
Optional reviewed explanation line per bundle; block, shortcode, cart and order received sections; Arabic translation
Frequently asked questions
Do I need an AI account?+
For the meaning similarity, yes: an OpenAI-compatible service with an embedding model (OpenAI, Azure OpenAI, OpenRouter, a local server such as Ollama...). Anthropic has no embedding endpoint. Without AI, the bought-together counts and the fallback by category and tag still work. You pay your provider; the plugin has no usage fees.
What does it cost to run?+
Each product is embedded once, then again only when its text changes; the Status tab estimates the tokens and the price before you start. Showing recommendations makes no AI request. The optional explanation lines cost one small chat request per bundle, written once. Your provider's prices are what count.
Is customer data sent to the AI or stored?+
No. The bought-together counts are computed on your site from the products of the orders; only product pairs and counts are stored. The AI receives product texts only.
Does it track visitors or set cookies?+
No. Suggestions depend on the products, not on the visitor, and the plugin sets no cookie. Adding to the cart uses WooCommerce's normal cart, which has its own cookies.
Does it give a discount for the bundle?+
No. The bundle uses the WooCommerce prices. If you want a discount for products bought together, use your pricing or discount plugin.
Does it work with HPOS?+
Yes. Orders are read through WooCommerce's order API, so HPOS and the legacy order storage both work; both are covered by the automated tests.
How big can my catalogue be?+
The similar products are computed in PHP in the background. Measured: about 20 seconds to 1.6 minutes of work for 1,000 products, and 29 minutes (vector size 256) to 4.5 hours (1536) for 10,000 products. Page views only read stored lists. For tens of thousands of products a vector database is the better tool.
What happens if the AI is down?+
Nothing visible: page views never call the AI. Indexing stops and continues later, and the stored lists, the counts and the fallback keep working.
Does it work in Arabic?+
Yes. The plugin is translated into Arabic and the admin works right to left. To match products written in different languages you need a multilingual embedding model.
More from TAQNIX
App templateNew
Taqnix POS Terminal – Offline-First Point of Sale PWA for Odoo 19: Next.js Source Code with Restaurant Tables, Kitchen Tickets, Loyalty and Arabic
A point of sale for Odoo 19 that runs in the browser as an installable app and keeps selling when the network is gone: every order is saved on the device first and sent to Odoo once, with the same call Odoo's own POS uses. Variants and combos, split payments, receipts with barcode and QR, restaurant tables and kitchen tickets, dashboard, stock lookup, layaway, loyalty, English and Arabic. Full Next.js source code; no Odoo module to install.
UAE E-Invoicing (PINT AE) for Odoo 19 Community – PINT AE XML for Your ASP, Readiness Checks, UBL XSD and Schematron Validation, E-Invoice Outbox
Prepare UAE e-invoices and credit notes in the PINT AE format (Peppol PINT AE Billing 1.0.4, UBL 2.1) for your Accredited Service Provider: TIN, TRN and tax categories, readiness checks per invoice and per batch, validation against the official XSD and Schematron, an outbox with statuses, a log and XML / ZIP export. It does not send anything. Arabic included.
AI Support Reply Drafts for WooCommerce – Order-Aware Customer Reply Drafts with Fact Checks
Draft replies to customer messages on the WooCommerce order screen with your AI model, from the order facts and your own shop texts only. Missing facts become [CHECK] placeholders; nothing is sent without a click and a confirmation.
AI Post and Product Translator for WordPress & WooCommerce – Linked Drafts, Glossary & hreflang
Translate posts, pages and WooCommerce products with your own AI into linked drafts: block-safe placeholders, a glossary and do-not-translate list, a structure check that refuses broken results, outdated detection, bulk runs with a cost estimate, a language switcher and hreflang links.
AI Support Reply Drafts for WooCommerce – Order-Aware Customer Reply Drafts with Fact Checks
Draft replies to customer messages on the WooCommerce order screen with your AI model, from the order facts and your own shop texts only. Missing facts become [CHECK] placeholders; nothing is sent without a click and a confirmation.
AI Post and Product Translator for WordPress & WooCommerce – Linked Drafts, Glossary & hreflang
Translate posts, pages and WooCommerce products with your own AI into linked drafts: block-safe placeholders, a glossary and do-not-translate list, a structure check that refuses broken results, outdated detection, bulk runs with a cost estimate, a language switcher and hreflang links.
AI Chatbot from Your Site Content for WordPress – Answers Only from Your Pages, with Sources
A chat widget that answers visitors only from your own pages, posts and products, links its sources, says so when the site has no answer and hands over to a person, with rate limits and spending caps.
AI Comment and Review Moderator – Spam, Abuse & Fake Review Filter for WordPress and WooCommerce
Score new comments and product reviews with your AI model, decide by your own thresholds, and review everything in one queue. Dry run first; nothing is ever deleted.
