LlamaIndex Vector Stores Elasticsearch

0.6.0 verified Mon Apr 27 auth: no python

Elasticsearch vector store integration for LlamaIndex. Provides ElasticsearchStore for indexing and querying vector embeddings with full Elasticsearch features. Current version: 0.6.0. Release cadence: follows llama-index releases.

pip install llama-index-vector-stores-elasticsearch

Common errors

error elasticsearch.BadRequestError: BadRequestError(400, 'x_content_parse_exception', 'failed to parse field [mappings.properties.embedding]') ↓

cause Elasticsearch index already exists with incorrect mapping for the embedding field (e.g., wrong dimension or type).

fix

Delete the existing index and let the store recreate it with correct mapping, or update the index mapping to match the expected dense_vector type.

error AttributeError: module 'llama_index.vector_stores.elasticsearch' has no attribute 'ElasticsearchStore' ↓

cause Old import path: from llama_index.vector_stores import ElasticsearchStore (incorrect).

fix

Use the correct import: from llama_index.vector_stores.elasticsearch import ElasticsearchStore

Warnings

breaking In version 0.6.0, the parameter `es_url` replaced the deprecated `host` and `port` parameters. Code using old parameters will fail. ↓

fix Use `es_url` instead of `host` and `port`.

deprecated The `distance_strategy` parameter is deprecated in favor of using `similarity` parameter with values like "cosine" or "l2". ↓

fix Replace `distance_strategy="COSINE"` with `similarity="cosine"`.

gotcha ElasticsearchStore does not automatically create the index if it doesn't exist. You must either pre-create the index with appropriate mappings or use the `create_index` parameter. ↓

fix Set `create_index=True` when initializing the store, or manually create the index with proper dense_vector field mapping.

gotcha When using authentication, ensure you pass both `es_user` and `es_password`. Passing only one may lead to unexpected connection errors. ↓

fix Always provide both user and password or use API key authentication.

Imports

ElasticsearchStore

from llama_index.vector_stores.elasticsearch import ElasticsearchStore

Official import path.

Quickstart

Initialize an ElasticsearchStore with environment variables for credentials.

import os
from llama_index.vector_stores.elasticsearch import ElasticsearchStore

vector_store = ElasticsearchStore(
    index_name="my_index",
    es_url=os.environ.get("ES_URL", "http://localhost:9200"),
    es_user=os.environ.get("ES_USER", "elastic"),
    es_password=os.environ.get("ES_PASSWORD", "changeme"),
)
print("Elasticsearch store created successfully.")