I'm using the client to retrieve product recommendations based on the bought-together and related-products models.
Some of our indexes don't have enough events to train models so they shouldn't be used for recommendations, but it's possible that we misconfigure this somewhere, or that the number of events drops for indexes that did have enough events at some point.
Where there's no trained model (but the index does exist) I get the following exception:
(stack trace limited to vendor namespace only)
Fatal error: Uncaught Algolia\AlgoliaSearch\Exceptions\NotFoundException: Index ai_recommend_webshop-products-se:related-products.indice.bin does not exist in ./vendor/algolia/algoliasearch-client-php/src/RetryStrategy/ApiWrapper.php:217
Stack trace:
#0 ./vendor/algolia/algoliasearch-client-php/src/RetryStrategy/ApiWrapper.php(167): Algolia\AlgoliaSearch\RetryStrategy\ApiWrapper->handleResponse()
#1 ./vendor/algolia/algoliasearch-client-php/src/RetryStrategy/ApiWrapper.php(101): Algolia\AlgoliaSearch\RetryStrategy\ApiWrapper->request()
#2 ./vendor/algolia/algoliasearch-client-php/src/RecommendClient.php(85): Algolia\AlgoliaSearch\RetryStrategy\ApiWrapper->write()
thrown in ./vendor/algolia/algoliasearch-client-php/src/RetryStrategy/ApiWrapper.php on line 217
I would like to catch this kind of exception, but it is identical in type to the exception I get when the index doesn't exist:
Fatal error: Uncaught Algolia\AlgoliaSearch\Exceptions\NotFoundException: Index does not exist in ./vendor/algolia/algoliasearch-client-php/src/RetryStrategy/ApiWrapper.php:217
Ideally this would be a different kind of exception so my code can handle missing/untrained models but doesn't handle missing indexes since that's a much more grave error.
Attempting to parse the exception message seems a little unreliable, as those messages might change without us noticing.
Steps To Reproduce
Request a recommendation from an existing index without a trained model.
Description
Hi there,
I'm using the client to retrieve product recommendations based on the
bought-togetherandrelated-productsmodels. Some of our indexes don't have enough events to train models so they shouldn't be used for recommendations, but it's possible that we misconfigure this somewhere, or that the number of events drops for indexes that did have enough events at some point.Where there's no trained model (but the index does exist) I get the following exception:
(stack trace limited to vendor namespace only)
I would like to catch this kind of exception, but it is identical in type to the exception I get when the index doesn't exist:
Ideally this would be a different kind of exception so my code can handle missing/untrained models but doesn't handle missing indexes since that's a much more grave error. Attempting to parse the exception message seems a little unreliable, as those messages might change without us noticing.
Steps To Reproduce