@dune/plugin-orama

Orama-backed search engine for Dune CMS sites.

Orama runs fully in-process — no external service or Docker container required. Swap it in for the built-in search engine when you need typo tolerance and better relevance scoring without the operational overhead of Meilisearch.

When to use

	Built-in	plugin-orama	plugin-meilisearch
External service	—	—	Required
Typo tolerance	✗	✓	✓
Relevance scoring	Basic	Good	Excellent
Index persistence across restarts	✗	Rebuilt on startup	✓
Suitable for	Small sites	Small–medium sites	Large sites

Installation

Add to your site's deno.json:

{
  "imports": {
    "@dune/plugin-orama": "jsr:@dune/plugin-orama"
  }
}

Usage

Enable from site.yaml:

plugins:
  - src: "jsr:@dune/plugin-orama"
    config:
      active: true         # default: true
      typoTolerance: true  # default: true (Orama tolerance: 1)
      excerptLength: 160   # default: 160 characters

That's it. Dune calls the engine's build() at startup and rebuild() on content changes. The same search API and templates work unchanged.

Configuration

plugins:
  - src: "jsr:@dune/plugin-orama"
    config:
      active: true
        # Set to false to register the engine without activating it.
        # Useful when running both built-in and Orama in parallel mode.

      typoTolerance: true
        # When true, searches tolerate one-character typos (Orama tolerance: 1).
        # Set to false for exact-match-only search.

      excerptLength: 160
        # Character length of excerpt returned per result.

Programmatic usage

import { createOramaEngine } from "@dune/plugin-orama/engine";

const engine = createOramaEngine(
  { tolerance: 1, excerptLength: 200 },
  pages,           // PageIndex[] from @dune/core
  { loadText },    // async (page) => string — body text loader
);

await engine.build();
const results = await engine.search("my query");
const suggestions = await engine.suggest("my");

Multi-engine setup

Register Orama alongside the built-in engine without activating it, then switch at runtime via the admin panel's Search tab:

plugins:
  - src: "jsr:@dune/plugin-orama"
    config:
      active: false   # register without activating

Switch the active engine at runtime via PATCH /admin/api/search/engines:

curl -X PATCH /admin/api/search/engines \
  -H "Content-Type: application/json" \
  -d '{"active": "orama"}'

Document schema

Each Dune page is indexed as:

{
  route: string;      // e.g. "/articles/my-post"
  title: string;
  body: string;       // plain text body loaded via loadText
  template: string;
  language: string;
  tags: string[];     // all taxonomy values flattened across vocabularies
}

The Orama index is rebuilt in memory on each startup. For large sites (thousands of pages) startup time increases proportionally to the body text loaded. Typical small-to-medium sites (< 1000 pages) build in under a second.

License

MIT