Elasticsearch / OpenSearch Adapter

A single adapter implementation that supports both Elasticsearch and OpenSearch. Select your engine at construction time.

When to Use

Large-scale search with millions of documents
Advanced relevance tuning with field boosts and custom analyzers
Existing Elasticsearch or OpenSearch infrastructure

Installation

npm install @searchfn/adapters

Usage

Elasticsearch

import { ElasticsearchAdapter } from "@searchfn/adapters";

const adapter = new ElasticsearchAdapter({
  node: "https://es.example.com:9200",
  engine: "elasticsearch",
  auth: { apiKey: process.env.ES_API_KEY! },
  indexPrefix: "myapp",
});

await adapter.initialize({
  resources: [{ name: "tasks", searchFields: ["title", "description"] }],
});

const ids = await adapter.search({ resource: "tasks", query: "docs" });

OpenSearch

import { OpenSearchAdapter } from "@searchfn/adapters";

const adapter = new OpenSearchAdapter({
  node: "https://opensearch.example.com:9200",
  auth: { username: "admin", password: process.env.OS_PASSWORD! },
  indexPrefix: "myapp",
});

OpenSearchAdapter is a convenience wrapper that sets engine: "opensearch" automatically.

Configuration

Option	Type	Default	Description
`node`	`string`	—	Required. Cluster endpoint URL.
`engine`	`"elasticsearch" \| "opensearch"`	`"elasticsearch"`	Select the engine dialect.
`auth`	`{ username, password } \| { apiKey }`	—	Cluster authentication credentials.
`indexPrefix`	`string`	`"searchfn"`	Prefix for index names (`{prefix}-{resource}`).
`requestTimeoutMs`	`number`	`30000`	Request timeout.
`retry.maxRetries`	`number`	`2`	Retry attempts for transient failures.
`retry.baseDelayMs`	`number`	`100`	Base delay between retries.
`retry.maxDelayMs`	`number`	`5000`	Maximum delay between retries.
`logger`	`AdapterLogger`	—	Optional adapter-level logger.

Capabilities

Capability	Supported
`persistent`	Yes
`searchAll`	Yes
`fuzzy`	Yes
`fieldBoosts`	Yes

Behavior

Single implementation handles both Elasticsearch and OpenSearch with dialect-aware query generation.
Index lifecycle (create, delete) is deterministic and idempotent.
Results are ordered by score with deterministic tie-breaking.
Unsupported engine values return DFQL_UNSUPPORTED.
Keep auth credentials in secure runtime secrets — never hardcode them.

Elasticsearch / OpenSearch Adapter

On this page