create_index_allmini.md

modify cluster settings

PUT _cluster/settings { "persistent" : { "plugins.ml_commons.only_run_on_ml_node" : false, "plugins.ml_commons.model_access_control_enabled": "true", "plugins.ml_commons.native_memory_threshold": "99"

} }

register model

retrieve model_id from the response

or retrieve task_id, look up it and finally retrieve model_id

POST /_plugins/_ml/models/_register { "name": "huggingface/sentence-transformers/all-MiniLM-L12-v2", "version": "1.0.1", "model_group_id": $model_goup_id, "model_format": "TORCH_SCRIPT" }

read response and copy task_id

GET /_plugins/_ml/tasks/INSERT_task_id

model_id =

deploy model using model_id

POST /_plugins/_ml/models/model_id/_deploy

creare pipiline con 2 mapping

PUT /_ingest/pipeline/all_mini_3_fields { "description": "all_mini_3_fields; Model: all-MiniLM-L12-v2", "processors": [ { "text_embedding": { "model_id": "mXgYKo8BlLTTsnWvtjbF", "field_map": { "text_en": "text_en_embedding", "category_en": "category_en_embedding", "text": "text_it_embedding" } } } ] }

creare indice con 2 embedding fields

PUT test_all_mini_cosine { "settings": { "index": { "knn": true, "default_pipeline": "all_mini_3_fields", "knn.algo_param.ef_search": 1000 } }, "mappings": { "properties": { "text_en_embedding": { "type": "knn_vector", "dimension": 384, "method": { "name": "hnsw", "space_type": "cosinesimil", "engine": "nmslib", "parameters": { "ef_construction": 1000, "m": 40 } } }, "text_it_embedding": { "type": "knn_vector", "dimension": 384, "method": { "name": "hnsw", "space_type": "cosinesimil", "engine": "nmslib", "parameters": { "ef_construction": 1000, "m": 40 } } }, "category_en_embedding": { "type": "knn_vector", "dimension": 384, "method": { "name": "hnsw", "space_type": "cosinesimil", "engine": "nmslib", "parameters": { "ef_construction": 1000, "m": 40 } } } } } }