Shoptimizer v1.7.0

New Features

• Added Korean language support. The language code is ko.

Bugfixes/Changes

• Improved the size mining regular expressions to cover edge cases.

Shoptimizer v1.6.1

New Features

• Added support for overriding any config file directly via the request by supplying a request header in the following format:

  Key: "{CONFIG FILE NAME PREFIX}_override"
  Value: [A Base64-encoded string of the config file's contents to override the matching config file]

  Example

  Key: "free_shipping_optimizer_config_override"
  Value: "ewogICJmcmVlX3NoaXBwaW5nX3BhdHRlcm5zIjogWwogIF0sCiAgInNoaXBwaW5nX2V4Y2x1c2lvbl9wYXR0ZXJucyI6IFsKICBdCn0K"

Bugfixes/Changes

• Introduced pre-sanitization of the request body in the form of converting numeric-type values of title and description attributes to their respective string representations to handle cases when those attributes are not string types.

Shoptimizer v1.6.0

New Features

• title-word-order-optimizer "WMM" tagging
  • When using the title-word-order-optimizer in tandem with a Word-Mix-Model-generated config file, products processed with this optimizer are applied a new tag called "WMM" to differentiate these from other optimized product tags.

Bugfixes/Changes

• pipenv has been replaced with requirements.txt. The CICD deployment script for GCP has been updated accordingly.
• Revamped the way that tagging is performed (string building instead of hardcoded).
• Added a .gitignore file.

Documentation

• Revamped the API Specification section to be broken down into sub-sections with simplified headers, and added TOC links to each.
• Updated the instructions on installation/unit testing due to the pip change.

Shoptimizer v1.5.0

New Features

• Added support for faster Cloud Build deployment times (caching)
  • Successive builds are now cached to reduce the deployment time in Cloud Build.
• Added French language support along with default FR config files.
• Added support for optional override of setting mining to be able to be toggled ON or OFF by setting the following HTTP request headers (set to "True" or "False"):
  • brand_mining_on, color_mining_on, gender_mining_on, and size_mining_on.
• Added support to allow optional override of controlling if mined product attributes are overwritten or not with their respective mined values using the following HTTP request headers (set to "True" or "False"):
  • color_mining_overwrite, gender_mining_overwrite, and size_mining_overwrite.

Bugfixes/Changes

• Enhanced clothing size regexes (for optimizers that use size mining) to decrease false positives while at the same time increasing match coverage, as well as cover "one-size-fits-all" cases for Japanese language.
• Upgraded Mecab version (to 1.0.4) and updated code and dependencies to be compatible with the latest version.
• Added precompiling of regexes for the title-word-order-optimizer to address performance bottlenecks.
• Refactored gender mining logic into its own module.

Documentation

• Clarified verbiage in the optimizer descriptions to more accurately describe the effect of each optimizer.
• Added explanations on the usage of Request Headers to control mining toggle settings.
• Added verbiage on how to clear the Cloud Build cache.

Shoptimizer v1.4.0-beta

New Features

Config Files

• Added default configuration files for German language.

Bugfixes/Changes

• Fixed a bug in the free-shipping-optimizer that involving unintended appending of a free shipping node to the product object by making the check for pre-existing shipping information case-insensitive.
• Modified the behavior of the title-word-order-optimizer to not identify a product as optimized unless words have been reordered. Previously, any changes to the title such as length truncation would have flagged this as an optimization for this optimizer.
• Due to removal of test keywords in the condition-optimizer sample configuration, tests have been updated accordingly for the condition-optimizer.

Config Files

• The title-word-order-optimizer configuration file has been refactored to contain two separate entries: "keyword_weights_by_gpc" (the original config contents) and "phrase_dictionary" (a list of terms that denote full multi-word phrases which should not be split up by the tokenizer). "phrase_dictionary" previously existed as a separate configuration file "title_word_order_dictionary.json" but has now been merged into a single configuration with "keyword_weights_by_gpc".

• The default Japanese-language configuration file for condition-optimizer has been updated to remove terms that cause false-positive matches.

Shoptimizer v1.3.1-beta

New Features

Shopping Exclusion Optimizer

• Added "Free Listing" to the list of surfaces that the Shopping Exclusion Optimizer excludes.

Title Word Order

• Makes sure phrases that are in the "dictionary" don't get split up so they can be detected as high-performing keywords.

Config Files

• Added language configs for Bahasa Indonesia.
• Added language configs for French for Free Shipping and Shopping Exclusion optimizations.
• Added additional promotional text patterns to EN configuration.

Bugfixes/Changes

• Added check and warning message if the Image Scoring Optimizer model cannot be found.

Shoptimizer v1.3.0-beta

New Features

Image Link Optimizer

• Added a new optimizer "image-link-optimizer" that adds support for automatically replacing images that are likely to be disapproved due to overlays or other image policy violations. The product must have alternate images populated in the additionalImageLinks field to use as replacements. This optimizer uses a Machine Learning model to determine which images are likely to be disapproved, so this optimizer requires more resources than others to function. See the optimizer documentation for details.

Config Files

• Added default French-language user-customizable configuration files.
• Added default English terms into the configuration file (promo_text_removal_optimizer_config_en.json) for promo-text-removal-optimizer.

Bugfixes/Changes

• title-word-order-optimizer: Added an option to append keywords to the back of the title instead of the front. See the optimizer documentation for details.
• Added recommended Optimizer ordering of the query parameters to the documentation.
• Add logic into title-length-optimizer that covers an edge case where the title and description match, but truncation of the title due to ellipsis results in them not matching anymore: when the title is equal to the description, the product should be skipped from being optimized, but was optimized anyway due to containing ellipsis which is sliced out and then compared against description. This has been fixed.
• Increased RAM requirements to 2GB for Cloud Run in the Cloud Build trigger configuration due to the increased demands of Image Link Optimizer.

Shoptimizer v1.2.0-beta

New Features

Multiple Optimizers

• Added support for using Google Product Category as number IDs instead of strings as the request input data.

Optional per-item exclusion attribute feature

• Added support for an optional "excludeOptimizers" attribute in each entry of an API request to control which optimizers should be skipped for that entry. See the Developer Guide for details.

Config Files

• Added additional default string patterns for the Japanese language.

Bugfixes/Changes

• In title-word-order-optimizer, changed the threshold number of characters for Japanese that controls words located "near the front of the title" from 12 to 18.
• Covered an edge case in title-length-optimizer where comparing the description with a truncated title that contains ellipsis resulted in unintentionally optimizing the product.

Misc

• Added a column "Potentially Modified Attributes" to the README which makes it transparent which Content API attributes could be modified by a specific optimizer.
• Fixed various unit test cases.

Shoptimizer v1.1.0-beta

New Features

title-word-order-optimizer

• Added support for a JSON-formatted configuration file called "title_word_order_options.json" for specifying options that controls the behavior of the title-word-order-optimizer. The options include descriptionIncluded and productTypesIncluded, boolean values that control whether or not the description and productTypes product attributes are checked for containing high-performing words. There is also an option called optimizationLevel which can be set to values "standard" or "aggressive" that controls whether or not to match product GPCs to the "title_word_order_config_{lang}.json" config's GPCs up to 3-levels deep (any GPCs with more depth are skipped from having the product optimized), or match the GPCs as-is for the aggressive mode.

• Added support for a config file called "title_word_order_blocklist_{lang}.json" which prevents detected high-performing words from being prepended to a title if they were found in this configuration file's word list.

• Added logic that prevents prepending high-performing words that are already situated near the front of the title. The exception is if those words get pushed out far enough by the prepend of other words not near the front of the title.

• Added a feature that ensures the title-word-order-optimizer always runs last no matter what order the optimizers are specified in the API query string params.

• Added support for specifying GPCs as ID numbers instead of their equivalent fully-qualified text strings. The full list of GPCs for any language can be found here.

• Added support for Japanese-language parsing.

• Added a default English config file.

• Added prevention of the prepending of detected high-performing keywords that are only a single character in length.

Bugfixes/Changes

title-word-order-optimizer

• Copy, instead of move, any detected high-performing keywords to the front of a product's title.

gtin-optimizer

• Removed support for this optimizer module.

Shoptimizer v1.0.0-beta

Initial beta release of base suite of optimizers for the Shoptimizer API.

See the README for installation and usage.