Remove all mentions of eis and gateway and deprecate flags that do #116692
Conversation
| @@ -15,6 +15,8 @@ | |||
| */ | |||
| public class ElasticInferenceServiceFeature { | |||
|
|
|||
| @Deprecated | |||
| public static final FeatureFlag ELASTIC_INFERENCE_SERVICE_FEATURE_FLAG = new FeatureFlag("eis"); | |||
There was a problem hiding this comment.
Maybe we could rename ELASTIC_INFERENCE_SERVICE_FEATURE_FLAG to DEPRECATED_INFERENCE_SERVICE_FEATURE_FLAG. Then it's also already clear from the constant name, which one to use.
| public static final FeatureFlag ELASTIC_INFERENCE_SERVICE_FEATURE_FLAG = new FeatureFlag("eis"); | ||
|
|
||
| public static final FeatureFlag INFERENCE_SERVICE_FEATURE_FLAG = new FeatureFlag("elastic_inference_service"); |
There was a problem hiding this comment.
Just wondering why we prefix the setting with ELASTIC_ and the feature flag not. Maybe this could lead to confusion considering that we also have generic interfaces named InferenceService.
WDYT?
There was a problem hiding this comment.
Ah, it was simply because the suggestion was Feature flag -Des.eis_feature_flag_enabled=true --> -Des.inference.elastic.feature_flag_enabled=true
Maybe we change the old one to DEPRCATRED_EIS_FEATURE_FLAG and the new one to ELASTIC_INFERENCE_SERVICE_FEATURE_FLAG? It's just used in tests and when "enabling" the flag. WDYT?
| @@ -105,9 +105,9 @@ private URI createUri() throws URISyntaxException { | |||
|
|
|||
| switch (modelId) { | |||
| case ElserModels.ELSER_V2_MODEL -> modelIdUriPath = "ELSERv2"; | |||
| default -> throw new IllegalArgumentException("Unsupported model for EIS [" + modelId + "]"); | |||
| default -> throw new IllegalArgumentException("Unsupported model for Elastic Inference Service [" + modelId + "]"); | |||
There was a problem hiding this comment.
Could we use a constant for Elastic Inference Service to make sure it stays consistent throughout log messages, error messages etc. I think we already have two places, where we use Elastic Inference Service (here and in the deprecation log warning message), right?
We could add it to ElasticInferenceService and maybe name it HUMAN_READABLE_NAME or something like that to distinguish from NAME?
There was a problem hiding this comment.
Good idea! Where should that live in your opinion? Added it to the InferencePlugin class. Lmk if thats ok.
There was a problem hiding this comment.
Thanks!
I think I would prefer if "human readable name" constant lives alongside the service provider NAME inside ElasticInferenceService.java. WDYT?
|
Hi @jaybcee, I've created a changelog YAML for you. |
|
Pinging @elastic/ml-core (Team:ML) |
💚 All backports created successfully
Questions ?Please refer to the Backport tool documentation |
…lastic#116692) (cherry picked from commit 0410229) # Conflicts: # x-pack/plugin/inference/qa/inference-service-tests/src/javaRestTest/java/org/elasticsearch/xpack/inference/InferenceCrudIT.java # x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/InferencePlugin.java # x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/elastic/ElasticInferenceServiceSparseEmbeddingsModel.java
… do (#116692) (#119943) * Remove all mentions of eis and gateway and deprecate flags that do (#116692) (cherry picked from commit 0410229) # Conflicts: # x-pack/plugin/inference/qa/inference-service-tests/src/javaRestTest/java/org/elasticsearch/xpack/inference/InferenceCrudIT.java # x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/InferencePlugin.java # x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/elastic/ElasticInferenceServiceSparseEmbeddingsModel.java * Fixing tests * Removing service api in 8.18 --------- Co-authored-by: Jason Botzas-Coluni <[email protected]>
We are deprecating the term
eisin lieu of Elastic Inference Service. We also want to stop exposing the implementation detail that this is a gateway. Deprecating 2 flags added 2 new ones and refactored some code to support this idea.Closes https://github.com/elastic/search-team/issues/8533