AI Visual Search & Shop the Look

Let shoppers search from a photo, find visually similar products and explore shoppable looks. Build product hotspots by hand or review AI suggestions before publishing. Support Luma and Hyvä with catalogue indexing and provider cost controls.

  • Magento 2.4.7 – 2.4.9
  • PHP 8.2 – 8.5
  • Version 1.0.3
$449
$314
Save 30%

One-off payment, with 12 months of updates.

What is included

  • 12 months of new versions and fixes; the versions released in that time stay yours
  • Install with Composer, or download a zip from your account
  • Licence for one production domain, staging and development copies included
  • 30-day money-back guarantee, refund policy

No subscription or automatic renewal. Keep using the versions included in your update period. Update and licence details

Composer package softaware/module-ai-visual-search

AI Visual Search & Shop the Look Regular price $449$314

Some shoppers know the look they want but cannot name it. AI Visual Search & Shop the Look gives them a photo-based route into your range, alternatives from a product they already like and curated scenes linked to actual catalogue items.

Search by camera or uploaded image

The storefront camera control accepts a supported photo from the camera or library. Similar-products mode compares it with indexed catalogue images and displays matches through your product cards. Category controls refine the selection without sending the photo again.

Find similar and the Visually similar product section offer another route, using existing catalogue vectors. Check real product suitability: colour and silhouette alone do not establish the same size, material or technical compatibility.

Publish a look your team has verified

Create a scene and add numbered product hotspots manually, or use Detect Items with AI to propose objects and catalogue matches. Confirm, swap or remove suggestions before they are shown publicly. Assign store views and insert the Shop the Look CMS widget; looks can also appear on related product pages.

Individual product controls let customers choose what to buy. Add all to cart adds supported items without required options; a sized or configurable product links to its page rather than guessing the customer's selection.

Customer-photo Shop the look is a separate mode: a vision model identifies items in the uploaded scene and presents matching catalogue candidates for each. Evaluate cluttered scenes separately from clean single-object photos.

Keep image discovery operationally manageable

Index the base image, small image or up to four gallery images per product. Status & Index shows coverage, queue, vector backend and a cost estimate. Changed images are re-embedded. The anonymous search log records result counts, first clicks and provider/budget problems.

Use similarity thresholds, visitor and website limits and AI Core budgets. Results depend on clear catalogue photography; indexing more poor images does not guarantee better matches.

Requirements and photo handling

Magento Open Source or Adobe Commerce 2.4.7–2.4.9, PHP 8.2–8.5 with the documented image extensions, SoftAware Core and AI Core. Luma and Hyvä are supported. Voyage AI provides image embeddings; item detection also needs a supported Anthropic or OpenAI vision model.

Uploaded photos are processed by those providers. EXIF is removed and images are re-encoded; local disk retention is normally off but can be enabled for troubleshooting. HEIC depends on browser conversion; JPEG, PNG and WebP are the documented formats.

Read the Visual Search and Shop the Look guide and test one photo, one similar-product journey and one reviewed look before launch.

See the workflow

Try the admin demo. Storefront examples are available on the Luma demo and Hyvä demo. Demonstration or offline-test data shows the workflow; evaluate AI matching and forecasts with your own representative catalogue.

Photo discovery

  • Camera or library upload on supported devices
  • Similar-products and customer-photo Shop-the-look modes
  • Store product cards with the relevant pricing context
  • Category narrowing without sending the image again
  • Supported image formats and configurable upload limits

Product alternatives

  • Find similar from the product gallery
  • On-demand Visually similar list
  • Existing catalogue vectors used for similar-product retrieval
  • Cached suggestions per product and customer group

Merchant looks

  • Create looks with store-view assignments
  • Manual product hotspots using name or SKU
  • Optional AI item detection and suggested matches
  • Confirm, swap or remove suggestions before publication
  • CMS widget and optional related looks on product pages
  • Add-all for items without options; option products open their pages

Index and operations

  • Base, small or up to four gallery images per product
  • Changed images re-indexed through AI Core
  • Coverage, queue, backend and cost estimate in Status & Index
  • Anonymous search log with results, clicks and problems
  • Similarity thresholds, visitor/site limits and shared budgets

Providers and processing

  • Voyage AI for image embeddings
  • Anthropic or OpenAI vision for item detection
  • EXIF stripping and re-encoding
  • No normal disk retention; optional troubleshooting retention
  • Photo processing by providers must be reflected in your deployment review

Questions before you install

Does image search identify the exact product?

It returns visual matches, not guaranteed identity or technical compatibility. Materials, dimensions and suitability still need the product information.

Are AI-created hotspots published immediately?

No. Item detection proposes matches and your team confirms, replaces or removes them before merchant-look suggestions are shown to shoppers.

Does Add all to cart choose sizes for customers?

No. Products requiring options open their product pages. The combined action adds supported products that do not need options.

Which providers do I need?

Voyage AI for image embeddings. AI item detection also requires a supported Anthropic or OpenAI vision model through AI Core.

Are uploaded photos saved?

Not normally. The module strips EXIF and re-encodes images, and sends them to configured providers for processing. Optional troubleshooting retention changes local storage behaviour.

Compatibility

Requirements and compatibility

Compatibility of AI Visual Search & Shop the Look
Magento2.4.7 – 2.4.9
PHP8.2 – 8.5
Latest version 1.0.3
composer.json requires php ~8.2.0||~8.3.0||~8.4.0||~8.5.0 ext-gd * ext-fileinfo * ext-json * magento/framework ~103.0.7 softaware/module-core ^1.0 softaware/module-ai-core ^1.2 magento/module-backend * magento/module-catalog * magento/module-catalog-inventory * magento/module-checkout * magento/module-cms * magento/module-config * magento/module-customer * magento/module-store * magento/module-ui * magento/module-widget *

Installation

Up and running in minutes

After you buy, create a Composer key in your account. Then, in the root of your Magento project:

  1. 01Add the repository and your key (once per project)

    composer config repositories.softaware composer https://repo.softawarecommerce.com
    composer config --auth http-basic.repo.softawarecommerce.com PUBLIC_KEY PRIVATE_KEY
  2. 02Install the module

    composer require softaware/module-ai-visual-search
  3. 03Enable it

    bin/magento setup:upgrade
    bin/magento setup:di:compile
    bin/magento setup:static-content:deploy
    bin/magento cache:flush

    The last three are only needed in production mode.

Prefer a zip? Every version you are entitled to can be downloaded from My modules. More about Composer access

User guide

How to set up and use AI Visual Search & Shop the Look

For version 1.0.3. The same guide comes with the module, in docs/user-guide.md.

1. Before you start

  1. Install the module (it brings SoftAware AI Core with it).
  2. Softaware > AI Services > AI Providers & Costs: add your Voyage AI key (product and photo images are embedded with Voyage AI). For "Detect Items with AI" and "Shop the look" with customer photos you also need an Anthropic or OpenAI key. Set a monthly budget there if you want a hard limit for all AI modules.
  3. Optional: if your store has more than about 5,000 products per store view, ask your host to install the OpenSearch k-NN plugin. AI Core then uses it automatically.

2. Switch it on

Softaware > AI Visual Search > Settings (or Stores > Configuration > Softaware > AI > AI Visual Search):

  • General > Enabled: Yes, per store view (switch the scope to the store view, untick "Use Default").
  • Leave the other settings as they are for a start.

Within 5 minutes the cron starts embedding the product images of the enabled store views. Watch the progress on Softaware > AI Visual Search > Status & Index: the coverage bar reaches 100 % when every product with an image is indexed. "Estimate Cost" shows what indexing costs before anything is sent (usually a few pence per 1,000 products). "Queue All Products" is only needed after you change the indexing settings.

3. What shoppers see

  • Camera button in the search box. Shoppers take or choose a photo, pick "Similar products" or "Shop the look", and get a results page. Category buttons above the results filter them without sending the photo again. The photo is not stored.
  • Find similar on the product image and a Visually similar list further down the product page.
  • Shop the Look wherever you place the widget, and on the product pages of the products in a look.

4. Create a look

  1. Softaware > AI Visual Search > Looks > Add New Look.
  2. Enter a title (and optionally an identifier and a short description), choose the store views, upload the image (JPEG, PNG or WebP, ideally 1600 to 2000 px wide) and click Save and Continue Edit.
  3. Add hotspots:
  • By hand: click on the image where the marker should be, then type a product name or SKU and pick the product. Drag markers to move them.
  • With AI: click Detect Items with AI. Each item found gets a dashed box and a suggested product with its similarity and other matches. Check each one: Confirm it, click one of the other matches, or Remove it. Suggestions are never shown to shoppers until you confirm them.
  1. Save Look. The grid shows how many hotspots and unchecked suggestions each look has.
  2. Show the look: Content > Pages (or Blocks/Widgets) > insert widget Shop the Look and choose the look. Product pages show it automatically when "Show Looks with This Product" is on.

"Add all to cart" adds every product that needs no options; products with sizes or colours link to their page.

5. Settings worth knowing

  • Minimum Similarity (Search by Photo): raise it if unrelated products appear, lower it if too few appear.
  • Images to Index: "Gallery Images" with 2 to 3 images per product helps when shoppers photograph products from other angles; each image is an extra embedding.
  • How images reach the AI: set in AI Providers & Costs > Embeddings > Send Images As (Automatic by default: links for public shops, image data for staging sites behind a password).
  • Limits and Privacy: searches per visitor per hour, searches per day for all visitors, and how long the anonymous search log is kept.

6. Search log

Softaware > AI Visual Search > Search Log lists every photo search: store view, type, number of products found, best similarity, the first product the shopper clicked and any problem (budget used up, key missing, AI service error). Searches without results tell you which products shoppers look for that you may not sell.

7. Troubleshooting

SymptomWhat to check
No camera buttonModule enabled for this store view? Camera Button setting? Full page cache refreshed?
Coverage stays at 0 %Voyage AI key, cron running, AI budget not used up (Status & Index shows all three)
"Search by photo is not available at the moment"Budget used up, key missing or daily limit reached (see Search Log "Problem")
No Find similar buttonThe product has no vector yet (not indexed, no image, or not visible in the catalogue)
Look not shownLook active and assigned to the store view, at least one confirmed hotspot, Shop the Look enabled

Changelog

Release notes

1.0.3

Latest
  • Settings moved back to the Softaware tab of Stores > Configuration, listed in the collapsible "AI" group (Softaware Core 1.2.0 section groups) instead of a separate "Softaware AI" tab. Requires softaware/module-ai-core ^1.2.

1.0.2

  • Settings moved to the new "Softaware AI" tab of Stores > Configuration (provided by AI Core 1.1.0), which groups the settings of all SoftAware AI modules. Requires softaware/module-ai-core ^1.1.

1.0.1

1.0.1 (2026-10-10)

Fixed

  • AI requests leave room for the model's thinking: current Claude models always think and count those tokens in the output limit, so an answer could be cut off ("Allow more output tokens") and the run was lost after it had been paid for. The output limit is only a ceiling; unused tokens are not charged.

1.0.0

1.0.0 - 2026-10-10

Added

  • Search by photo: camera button in the search box (Luma and Hyvä, camera or photo library on phones), in-browser downscaling, server-side validation (finfo + getimagesize, size and pixel limits, EXIF removed), Voyage AI image embedding through AI Core, results page with category filter, customer group prices and add to cart.
  • Find similar on product pages (stored vectors, no AI request at view time, full-page-cache friendly fragment).
  • Shop the Look: merchant looks with hotspots (admin editor, CMS widget, product page lookbook, add to cart per item and "Add all to cart"), AI item detection with a vision model as suggestions the merchant confirms, and "Shop the look" for customer photos.
  • AI Core vector profile visual_search (base, small or up to 4 gallery images per product; public URL or inline image data), Status & Index page with coverage, backend, queue, budget and cost estimate.
  • Anonymous search log, hourly limit per visitor, daily limit per website, AI Core budget guard.
  • CLI softaware:visual-search:status and softaware:visual-search:search; hourly cleanup cron.
  • Developer test doubles (hidden setting softaware_visual_search/developer/fake_ai) and unit tests.
  • Built on the shared AiCore APIs: one profile with several vectors per product (replaces the extra profiles visual_search_2..4), "Find similar" via ProductVectorIndexInterface::searchByEntity() (the vector copy table is removed), photo search via searchByVector() and no-retry query embeddings, product images as AiCore media files (image delivery setting moved to AiCore), image processing via ImageProcessorInterface, rate limits via RateLimiterInterface, index status and estimate via VectorIndexStatusInterface, AI cost panel on Status & Index, item detection double as an AiCore offline responder (old developer settings migrated).

Fixed (pre-release review)

  • Products that are not in a category (or not yet in the category index) were missing from photo search results, Find similar and Shop the Look: visibility and status are now filtered by attribute instead of the root category's index.
  • "Add all to cart" with only unavailable products (out of stock, disabled, not in the website) now shows a message instead of silently redirecting to the cart; unexpected add-to-cart errors are reported too.
  • The heading level of a look (widget parameter heading_level) is limited to h1 to h6.
  • The look editor shows product names as plain text (sample names with ™ were shown literally).
  • Unit tests run with the module's own phpunit.xml.dist.

From the blog

Guides and articles