eland

Elastic’s Python client for loading and managing machine-learning models in Elasticsearch. A pandas-like DataFrame API over Elasticsearch indices, plus helpers for importing trained models for in-cluster inference.

• Docs: https://eland.readthedocs.io/
• GitHub: https://github.com/elastic/eland

Role in LTR

In native Elasticsearch LTR (8.12+), the GBDT model is trained outside the cluster — typically an XGBoost XGBRanker (LambdaMART). eland provides the helpers that serialize that trained ranker and upload it into Elasticsearch as a deployed model, where it is then applied by the learning_to_rank rescorer.

Related

• Elasticsearch · XGBoost · Elasticsearch Learning to Rank · LambdaMART

Articles

• Learning To Rank (LTR) - Elasticsearch Native — names eland as the model-upload path