Adapters

SearchFn ships five adapters that all implement the SearchAdapter contract. Two use SearchFn's built-in search engine (tokenization, BM25 scoring, inverted index, fuzzy matching). Three delegate to external search backends. Swap between them without changing your application code.

Adapter Comparison

Built-in Engine

These adapters run SearchFn's own search engine in-process. No external services required.

Adapter	Environment	Persistent	Fuzzy	Field Boosts	Best for
MemoryAdapter	Browser / Node	No	Yes	Yes	In-process server search, tests, CLIs, small-to-moderate datasets
IndexedDbAdapter	Browser	Yes	Yes	Yes	Offline-first browser apps, PWAs, client-side search

External Backends

These adapters are thin wrappers that translate the SearchAdapter contract into the backend's native API.

Adapter	Environment	Persistent	Fuzzy	Field Boosts	Best for
PostgresAdapter	Server	Yes	Yes	No	Production apps already using Postgres
MeilisearchAdapter	Server	Yes	Yes	No	Dedicated search with typo tolerance and faceting
Elasticsearch/OpenSearch	Server	Yes	Yes	Yes	Large-scale search with advanced relevance tuning

Choosing an Adapter

MemoryAdapter runs the full search engine in-process with zero setup. Use it for server-side search where the dataset fits in memory, CLI tools, tests, and any environment where you want search without external dependencies.

IndexedDbAdapter is for browser and offline-first apps. It uses the same built-in engine as MemoryAdapter, but persists the inverted index in IndexedDB with LRU caches — search survives page reloads and works offline.

Both built-in engine adapters are production-ready. Use an external backend when your use case specifically needs:

Shared search state across multiple clients or servers
Datasets that exceed process or browser memory
Backend-specific features (Postgres tsvector tuning, Meilisearch facets, Elasticsearch analyzers)

PostgresAdapter is the simplest external option if you already run Postgres. It uses native tsvector full-text search with no additional infrastructure.

MeilisearchAdapter gives you a dedicated search engine with built-in typo tolerance and fast indexing. Good for user-facing search with high relevance requirements.

Elasticsearch/OpenSearch Adapter is for large-scale deployments that need advanced relevance tuning, field boosts, and cluster scalability.

Common Interface

All adapters — built-in and external — support the same operations:

initialize — declare resources and searchable fields
index — batch index documents
search — search a single resource by query
remove — remove documents by ID
clear — remove all documents from a resource
dispose — release resources and connections

Most adapters also support searchAll (cross-resource search) and fuzzy matching. See individual adapter pages for details.

Lifecycle

Every adapter follows the same lifecycle:

Construct — create the adapter with configuration.
Initialize — call initialize() to declare resources and search fields.
Use — call index, search, searchAll, remove, clear.
Dispose — call dispose() to release resources. After disposal, call initialize() again before further operations.

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

const adapter = new MemoryAdapter();

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

await adapter.index({
  resource: "tasks",
  documents: [{ id: "1", fields: { title: "Hello", description: "World" } }],
});

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

await adapter.dispose();