Elasticsearch 임베딩

Elasticsearch 임베딩

Elasticsearch의 임베딩 모델을 LlamaIndex에서 사용하는 예시예요. Elasticsearch 임베딩과 벡터 스토어를 함께 설정해서 인덱스를 만들고 쿼리하는 방법을 보여줘요.

출처: 문서

본문

colab에서 이 노트북을 열고 있다면 LlamaIndex 🦙를 설치해야 할 거예요.

%pip install llama-index-vector-stores-elasticsearch
%pip install llama-index-embeddings-elasticsearch

!pip install llama-index

# imports


from llama_index.embeddings.elasticsearch import ElasticsearchEmbedding
from llama_index.vector_stores.elasticsearch import ElasticsearchStore
from llama_index.core import StorageContext, VectorStoreIndex
from llama_index.core import Settings

# get credentials and create embeddings


import os


host = os.environ.get("ES_HOST", "localhost:9200")
username = os.environ.get("ES_USERNAME", "elastic")
password = os.environ.get("ES_PASSWORD", "changeme")
index_name = os.environ.get("INDEX_NAME", "your-index-name")
model_id = os.environ.get("MODEL_ID", "your-model-id")




embeddings = ElasticsearchEmbedding.from_credentials(
    model_id=model_id, es_url=host, es_username=username, es_password=password
)

# set global settings
Settings.embed_model = embeddings
Settings.chunk_size = 512

# usage with elasticsearch vector store


vector_store = ElasticsearchStore(
    index_name=index_name, es_url=host, es_user=username, es_password=password
)


storage_context = StorageContext.from_defaults(vector_store=vector_store)


index = VectorStoreIndex.from_vector_store(
    vector_store=vector_store,
    storage_context=storage_context,
)


query_engine = index.as_query_engine()




response = query_engine.query("hello world")