Search and CMS integrations

Point the search calls you already make at DocSlurp. Get cited passages back.

Keep your search UI and swap the engine underneath. Point existing Algolia, Pagefind, Meilisearch or Typesense calls at DocSlurp and get cited passages from PDFs, Office files and media without rewriting the interface. Or add a search box to Astro, Next.js, WordPress or any HTML page.

Keep your search client

Algolia

PfPagefind

Meilisearch

Typesense

Pricing FAQ

Annual plans are billed up front.

Refunds are prorated for the unused months.

Seats can be added at any time.

Your workspace

Ship it on your site

Astro

Next.js

WordPress

</>Plain HTML

On this page

Already on a search service?

AlgoliaKeep conventional v4/v5 search calls and connect your existing search interface.PagefindKeep search(term), result metadata and the lazy data() interface in custom renderers.MeilisearchPreserve index(id).search() and familiar hit and pagination envelopes.TypesenseUse collection search and ordered multi-search with nested document hits.

Adding search to a site?

Astro ComponentA native Astro component with a small framework-free browser widget.Next.js RecipeRegister a custom element from a client component, with SSR-safe imports.WordPress PluginInstall a shortcode with a server-side proxy that keeps the server key private.HTML / other CMS Custom elementA custom element for sites with their own asset pipeline and templates.

How it works

Your search interfaceInstantSearch, a Pagefind renderer, a custom box. Unchanged.
Compatibility adapterSpeaks the original client’s request and response shapes.
DocSlurp retrievalHybrid search across every extracted page, with citations and usage.

What changes

The client constructor and key. Index or collection names become a DocSlurp workspace: a slug like product-docs, a UUID, or default.

What stays

Your markup, routing, keyboard behavior and hit rendering. Response envelopes keep familiar fields such as hits, found and data().

What’s not emulated

Indexing, settings and admin APIs, ranking rules, analytics and static index files. Content goes in through native ingestion, and results are passages rather than whole pages.

Before you start

Create a browser key

On API keys, create a browser key with the search scope and add your site to its allowed websites.

Upload and wait

Upload your content with the SDK, CLI or HTTP API and wait for the run to finish. Adapters only search; they can’t upload.

Swap the client

Replace one constructor. Keep the rest of your search code, then run the go-live checklist.

Node SDKCLI

ingest.mjs

import { createDocSlurpClient } from '@docslurp/sdk';

// Server-side only: DOCSLURP_API_KEY is a server key with the ingest scope.
const client = createDocSlurpClient();

const accepted = await client.upload({
  urls: ['https://example.com/docs/install', 'https://example.com/docs/billing'],
  workspaceId: 'product-docs', // omit to use your organization's default workspace
});
await client.waitForRun(accepted.run.id, { timeoutMs: 600_000 });

Terminal

# Upload local files and wait for indexing to finish.
docslurp upload ./site-export/*.pdf --workspace product-docs --watch

# Check that search returns results before you switch traffic.
docslurp search 'how do I install' --workspace product-docs --json

Re-ingesting a source with the same canonicalId replaces it without a search gap. See the SDK reference for uploads, URL ingestion and replacements.

Compatibility matrix

Unsupported options fail loudly with HTTP 400 instead of being silently dropped, so a filter you rely on can never quietly disappear.

Capability	Algolia	Pagefind	Meilisearch	Typesense
Keyword query	Supportedquery	Supportedsearch(term)	Supportedq	Supportedq + query_by
Pagination	Supportedpage, hitsPerPage	Supportedall results, lazy data()	Supportedoffset/limit, page	Supportedpage, per_page
Retrieval window	Supported≤ 1,000 hits	Partial100 passages	Partial100 passages	Partial100 passages
Batched queries	Supported≤ 10 requests	Not applicable—	Not supportedmulti-search	SupportedmultiSearch, ≤ 10
Filters	Partialdocumented subset	Not supported400	Not supported400	Not supported400
Facets	Partialwindow-scoped	Not supportedfilters() returns {}	Not supported400	Not supportedfacet_counts: []
Custom sort	Not supportedrelevance only	Not supported400	Not supported400	Not supported400
Highlighting	Partialpre/post tags	Not supportedescaped excerpt	Not supported—	Not supported—
Index & settings APIs	Partialmigration helpers	Not supporteduse native ingest	Not supporteduse native ingest	Not supporteduse native ingest

Keys and security

Key	Prefix	Lives in	Use it for
Browser key search-only, allowed websites only	`dslp_pub_`	Browser bundles, `PUBLIC_` / `NEXT_PUBLIC_`	Adapters, widget and Astro component on deliberately public content
Search-only server key	`dslp_priv_`	Server secrets, WordPress options	Proxies: WordPress plugin, Next.js route handler
Server key with ingest	`dslp_priv_`	CI or backend secret store	Uploads, URL ingestion and run waiting. Never in a browser or an agent prompt.

The prefix makes a leaked key easy to spot and classify. Every SDK surface that can run in a browser throws before sending a dslp_priv_ key, and a key’s type can’t be changed after creation. Every adapter request keeps DocSlurp’s authorization, allowed-website checks, rate limits and usage accounting. Meilisearch and Pagefind use Authorization: Bearer; Typesense also accepts X-TYPESENSE-API-KEY and Algolia X-Algolia-API-Key. Responses are sent with Cache-Control: private, no-store.

Limits and errors

Behavior	Contract
Retrieval window	Meilisearch, Typesense and Pagefind rank the first 100 passages per query. Pages that would reach past it return 400. `docslurp.bounded`, `windowLimit` and `countIsExact: false` mark every count as a window count.
Page size	Default 20 (Meilisearch, Typesense); Pagefind returns the whole window and loads `data()` lazily.
Batches	Up to 10 queries for Typesense multi-search and Algolia batches.
Empty query	Returns zero results without searching. It never lists the workspace’s contents.
Wildcard `*`	Rejected. List documents with the native API.
Request size	Queries up to 16,000 characters; request bodies up to 64 KB.
SDK transport	30-second deadline per request. Non-2xx responses throw with the status and body.
Ranking	DocSlurp hybrid retrieval with query expansion, reranking and other LLM-backed enhancements disabled, keeping adapter calls fast and inexpensive.

Unsupported optionHTTP

HTTP/1.1 400 Bad Request
Content-Type: application/json

{
  "code": "UNSUPPORTED_SEARCH_OPTION",
  "error": "meilisearch: filter is not supported. Use the native search API for retrieval controls."
}

400: UNSUPPORTED_SEARCH_OPTION: an option outside the contract, or invalid JSON.
401: Missing Authorization header (or provider key header).
403: The key lacks the search scope, is invalid for this origin, or hit its rate limit.
404: The workspace doesn’t exist for this key’s organization.

Troubleshooting

Cannot find module “@docslurp/sdk/search-adapters”

Your installed SDK is older than 0.2.0, which added these exports. Run npm install @docslurp/sdk@latest, then pin the version you test.

Algolia returns zero hits for every query

Batched Algolia requests return empty results for an index name that doesn’t match a workspace, rather than an error. Check the slug or UUID in the workspace settings, or use default.

Browser request fails with CORS or 403

Use a browser key with the search scope, add your site under allowed websites, and confirm the key’s organization owns the workspace. WordPress keeps its server key behind its proxy.

Results are empty right after upload

Upload is accepted before it is indexed. Wait for waitForRun() or docslurp upload --watch, check docslurp.availability, and remember that empty queries never return results.

HTTP 400 UNSUPPORTED_SEARCH_OPTION

The request used an option outside the supported contract: filters, facets, sort, wildcard *, a query_by other than text, or a page beyond the first 100 passages. Remove it or move that query to the native search API.

Source links are missing or wrong

Results carry a citation.documentId, not your site’s URL. Map document IDs to real public routes in resolveUrl. Pagefind rejects missing or unsafe URLs instead of guessing.

WordPress returns 429, 502 or 503

The proxy rate cap was hit, DocSlurp was unreachable, or plugin settings are incomplete. Check Settings > DocSlurp Search and your server logs, without logging the key.

Ranking differs from the old engine

DocSlurp ranks passages with its own hybrid retrieval. Compare representative queries and tune your content or the native retrieval controls; don’t expect identical order.

Examples

packages/sdk/examples/search-adapters.mjs runs every adapter against a real, already-indexed workspace. It sends four retrieval requests, which may incur embedding usage, so don’t wire it into an automatic smoke test.

Run the adapter exampleShell

pnpm --filter @docslurp/sdk build
DOCSLURP_API_KEY=… DOCSLURP_WORKSPACE_ID=product-docs \
  node packages/sdk/examples/search-adapters.mjs "installation"

Verify a local build before release

Build, test and packShell

pnpm --filter @docslurp/sdk build
pnpm --filter @docslurp/contrib-search-compat build
pnpm --filter @docslurp/sdk test
node scripts/package-wordpress-search.mjs --check
pnpm --filter @docslurp/sdk pack --pack-destination /tmp

The packed SDK must include dist/search-adapters.js, its declarations, dist/search-widget.js and integrations/astro/DocSlurpSearch.astro. Release steps live in docs/releases.md; don’t assume a checkout change is already on npm.

Next steps

Search referenceNative retrieval controls: filters, reranking, hybrid weights and response contracts.Embeddable searchTry the live widget and theme it for your site.Integration roadmapPlanned CMS connectors, documentation-site plugins and language SDKs.

Adapters were checked against the official Algolia client, Pagefind API, Meilisearch search API and Typesense search API. The supported subsets above are DocSlurp’s contract; those services evolve independently.