This package makes it easy to add full text search support to your models with Laravel 12 and 13.
Important
The features from the Scout driver in this repo have been merged upstream into Laravel Scout natively.
So we've temporarily paused development in this repo and plan to instead address any issues or improvements in the native Laravel Scout driver instead.
If there are any Typesense-specific features that would be hard to implement in Laravel Scout natively (since we need to maintain consistency with all the other drivers), then at that point we plan to add those features into this driver and maintain it as a "Scout Extended Driver" of sorts. But it's too early to tell if we'd want to do this, so we're in a holding pattern on this repo for now.
In the meantime, we recommend switching to the native Laravel Scout driver and report any issues in the Laravel Scout repo.
- Installation
- Usage
- Standalone Typesense (without Scout)
- Migrating from siberfx/typesense-scout
- Authors
- License
The Typesense PHP SDK uses httplug to interface with various PHP HTTP libraries through a single API.
First, install the correct httplug adapter based on your guzzlehttp/guzzle version. For example, if you're on
Laravel 8, which includes Guzzle 7, then run this:
composer require php-http/guzzle7-adapterThen install the driver:
composer require siberfx/typesense-scoutAnd add the service provider:
// config/app.php
'providers' => [
// ...
Siberfx\Typesense\TypesenseServiceProvider::class,
],Ensure you have Laravel Scout as a provider too otherwise you will get an "unresolvable dependency" error
// config/app.php
'providers' => [
// ...
Laravel\Scout\ScoutServiceProvider::class,
],Add SCOUT_DRIVER=typesense to your .env file
Then you should publish scout.php configuration file to your config directory
php artisan vendor:publish --provider="Laravel\Scout\ScoutServiceProvider"In your config/scout.php add:
'typesense' => [
'api_key' => 'abcd',
'nodes' => [
[
'host' => 'localhost',
'port' => '8108',
'path' => '',
'protocol' => 'http',
],
],
'nearest_node' => [
'host' => 'localhost',
'port' => '8108',
'path' => '',
'protocol' => 'http',
],
'connection_timeout_seconds' => 2,
'healthcheck_interval_seconds' => 30,
'num_retries' => 3,
'retry_interval_seconds' => 1,
],If you are unfamiliar with Laravel Scout, we suggest reading it's documentation first.
After you have installed scout and the Typesense driver, you need to add the
Searchable trait to your models that you want to make searchable. Additionaly,
define the fields you want to make searchable by defining the toSearchableArray method on the model and implement TypesenseSearch:
<?php
namespace App;
use Illuminate\Database\Eloquent\Model;
use Siberfx\Typesense\Interfaces\TypesenseDocument;
use Laravel\Scout\Searchable;
class Todo extends Model implements TypesenseDocument
{
use Searchable;
/**
* Get the indexable data array for the model.
*
* @return array
*/
public function toSearchableArray()
{
return array_merge(
$this->toArray(),
[
// Cast id to string and turn created_at into an int32 timestamp
// in order to maintain compatibility with the Typesense index definition below
'id' => (string) $this->id,
'created_at' => $this->created_at->timestamp,
]
);
}
/**
* The Typesense schema to be created.
*
* @return array
*/
public function getCollectionSchema(): array {
return [
'name' => $this->searchableAs(),
'fields' => [
[
'name' => 'id',
'type' => 'string',
],
[
'name' => 'name',
'type' => 'string',
],
[
'name' => 'created_at',
'type' => 'int64',
],
],
'default_sorting_field' => 'created_at',
];
}
/**
* The fields to be queried against. See https://typesense.org/docs/0.24.0/api/search.html.
*
* @return array
*/
public function typesenseQueryBy(): array {
return [
'name',
];
}
}Then, sync the data with the search service like:
php artisan scout:import App\\Models\\Todo
After that you can search your models with:
Todo::search('Test')->get();
The searchable() method will chunk the results of the query and add the records to your search index. Examples:
$todo = Todo::find(1);
$todo->searchable();
$todos = Todo::where('created_at', '<', now())->get();
$todos->searchable();You can send multiple search requests in a single HTTP request, using the Multi-Search feature.
$searchRequests = [
[
'collection' => 'todo',
'q' => 'todo'
],
[
'collection' => 'todo',
'q' => 'foo'
]
];
Todo::search('')->searchMulti($searchRequests)->paginateRaw();You can generate scoped search API keys that have embedded search parameters in them. This is useful in a few different scenarios:
- You can index data from multiple users/customers in a single Typesense collection (aka multi-tenancy) and create scoped search keys with embedded
filter_byparameters that only allow users access to their own subset of data. - You can embed any search parameters (for eg:
exclude_fieldsorlimit_hits) to prevent users from being able to modify it client-side.
When you use these scoped search keys in a search API call, the parameters you embedded in them will be automatically applied by Typesense and users will not be able to override them.
<?php
namespace App;
use Illuminate\Database\Eloquent\Model;
use Laravel\Scout\Searchable;
use Siberfx\Typesense\Concerns\HasScopedApiKey;
use Siberfx\Typesense\Interfaces\TypesenseDocument;
class Todo extends Model implements TypesenseDocument
{
use Searchable, HasScopedApiKey;
}Generate a scoped search key from a parent search-only API key, embedding the
parameters you want to enforce (e.g. a per-tenant filter_by and/or an
expires_at), then use it for searching:
// Generate a scoped key that locks searches to a single tenant.
$scopedKey = Todo::generateScopedSearchKey('parent-search-only-key', [
'filter_by' => 'company_id:42',
'expires_at' => now()->addHour()->timestamp,
]);
// Use a (pre-)generated scoped key for subsequent searches.
Todo::setScopedApiKey($scopedKey)->search('todo')->get();You can also create new API keys server-side:
app(\Siberfx\Typesense\Typesense::class)->createApiKey([
'description' => 'Search-only key',
'actions' => ['documents:search'],
'collections' => ['*'],
]);Standard Scout where, whereIn and whereNotIn clauses are supported:
Todo::search('shoes')
->where('user_id', 1)
->whereIn('status', ['open', 'in_progress'])
->whereNotIn('team_id', [9, 10])
->get();For comparison and range filters, pass the operator as an array value:
Todo::search('shoes')
->where('price', ['>', 100]) // price:>100
->where('rating', ['[3..5]']) // rating:[3..5]
->get();Boolean values are rendered as Typesense literals (true/false) automatically.
Search a vector field by nearest neighbours. Use nearestNeighbors() to build
the vector_query for you:
// Pure vector search: query '*' plus a vector field.
Todo::search('*')
->nearestNeighbors('embedding', [0.12, 0.34, 0.56], k: 10)
->get();
// Hybrid search: combine a text query with a vector query.
Todo::search('running shoes')
->nearestNeighbors('embedding', $vector, k: 20, distanceThreshold: 0.3, alpha: 0.4)
->get();Or pass a raw vector_query string for full control:
Todo::search('*')
->vectorQuery('embedding:([0.12, 0.34, 0.56], k:10, alpha:0.4)')
->get();Collection-level admin operations are available on the Typesense instance
(resolve it from the container or via the Typesense facade).
use Siberfx\Typesense\Typesense;
$typesense = app(Typesense::class);
// Synonyms (per collection)
$typesense->upsertSynonym('todos', 'coat-synonyms', [
'synonyms' => ['blazer', 'coat', 'jacket'],
]);
$typesense->retrieveSynonyms('todos');
$typesense->retrieveSynonym('todos', 'coat-synonyms');
$typesense->deleteSynonym('todos', 'coat-synonyms');
// Curation / overrides (per collection)
$typesense->upsertOverride('todos', 'promote-tidy', [
'rule' => ['query' => 'tidy', 'match' => 'exact'],
'includes' => [['id' => '123', 'position' => 1]],
]);
$typesense->retrieveOverrides('todos');
$typesense->retrieveOverride('todos', 'promote-tidy');
$typesense->deleteOverride('todos', 'promote-tidy');
// Collection aliases
$typesense->upsertAlias('todos', ['collection_name' => 'todos_v2']);
$typesense->retrieveAliases();
$typesense->retrieveAlias('todos');
$typesense->deleteAlias('todos');
// Analytics rules
$typesense->upsertAnalyticsRule('popular-todos', [
'type' => 'popular_queries',
'params' => [/* ... */],
]);
$typesense->retrieveAnalyticsRules();
$typesense->retrieveAnalyticsRule('popular-todos');
$typesense->deleteAnalyticsRule('popular-todos');use Siberfx\Typesense\Typesense;
$typesense = app(Typesense::class);
// Search presets
$typesense->upsertPreset('listing', ['value' => ['query_by' => 'name']]);
$typesense->retrievePresets();
$typesense->retrievePreset('listing');
$typesense->deletePreset('listing');
// Stopwords
$typesense->upsertStopword('common', ['stopwords' => ['a', 'the'], 'locale' => 'en']);
$typesense->retrieveStopwords();
$typesense->retrieveStopword('common');
$typesense->deleteStopword('common');
// Stemming dictionaries (no delete endpoint)
$typesense->upsertStemmingDictionary('plurals', [['word' => 'people', 'root' => 'person']]);
$typesense->retrieveStemmingDictionaries();
$typesense->retrieveStemmingDictionary('plurals');
// Conversation models (conversational / RAG search)
$typesense->createConversationModel([
'model_name' => 'openai/gpt-3.5-turbo',
'api_key' => 'OPENAI_API_KEY',
'history_collection' => 'conversation_store',
]);
$typesense->retrieveConversationModels();
$typesense->retrieveConversationModel('model-id');
$typesense->updateConversationModel('model-id', ['system_prompt' => '...']);
$typesense->deleteConversationModel('model-id');
// Natural language search models
$typesense->createNLSearchModel(['model_name' => 'openai/gpt-4', 'api_key' => '...']);
$typesense->retrieveNLSearchModels();
$typesense->retrieveNLSearchModel('nl-model-id');
$typesense->updateNLSearchModel('nl-model-id', [/* ... */]);
$typesense->deleteNLSearchModel('nl-model-id');Love using Scout with your Eloquent models, but every now and then you just want
to talk to Typesense directly? Maybe you're indexing data that isn't a model,
building a collection schema by hand, running a federated multi_search, or
reaching an admin API (keys, aliases, presets, analytics…) that Scout doesn't
expose. That's exactly what this is for.
Meet TypesenseDirect — a friendly, Scout-free way to use the raw Typesense
client, with a clean helper API on top and multi-cluster support built in.
Note
This lives happily next to the Scout driver and shares nothing with it at
runtime. Reaching for TypesenseDirect never changes how your models'
Model::search() behaves — use whichever fits the moment.
Which one do I want?
| You want to… | Use |
|---|---|
| Search your Eloquent models the Laravel way | Scout (Model::search(...)) |
| Index/search data that isn't a model, or build schemas by hand | TypesenseDirect |
Bulk-import, federated multi_search, or talk to a 2nd cluster |
TypesenseDirect |
| Manage keys / aliases / presets / analytics / stopwords | TypesenseDirect |
Drop down to the raw \Typesense\Client |
TypesenseDirect::connection()->client() |
Already using the Scout driver? Then there's nothing to configure — the
default connection reuses your existing Typesense credentials. Copy this into a
route, tinker, or a command and run it:
// 1. Create a collection
TypesenseDirect::ensureCollection([
'name' => 'books',
'fields' => [
['name' => 'title', 'type' => 'string'],
['name' => 'year', 'type' => 'int32'],
],
'default_sorting_field' => 'year',
]);
// 2. Add some documents (bulk)
TypesenseDirect::documents('books')->import([
['id' => '1', 'title' => 'Dune', 'year' => 1965],
['id' => '2', 'title' => 'Neuromancer', 'year' => 1984],
], 'upsert');
// 3. Search 🎉
$results = TypesenseDirect::search('books', [
'q' => 'dune',
'query_by' => 'title',
]);
echo $results['found']; // 1
echo $results['hits'][0]['document']['title']; // "Dune"That's the whole loop: create → import → search. Everything below is just more of the same surface, one topic at a time.
Nothing is required to get started: the default connection inherits
scout.typesense.client-settings, so any app already using the Scout driver has
standalone access immediately. Publish the config only when you want to define
extra connections or override values:
php artisan vendor:publish --tag=typesense-config// config/typesense.php
return [
'default' => env('TYPESENSE_CONNECTION', 'default'),
'connections' => [
// Inherits scout.typesense.client-settings for anything left null.
'default' => [
'api_key' => env('TYPESENSE_API_KEY'),
'nodes' => [ /* host/port/protocol … */ ],
'nearest_node' => null,
// timeouts/retries default to null -> inherit scout settings
],
// A second, fully-specified cluster.
'analytics' => [
'api_key' => env('TYPESENSE_ANALYTICS_KEY'),
'nodes' => [
['host' => 'analytics.example.com', 'port' => '443', 'protocol' => 'https'],
],
],
],
];Tip
The Scout fallback applies only to the default connection, so existing
apps get standalone access for free. Named connections (like analytics
above) are fully independent and must be spelled out completely.
Pick whatever reads best where you are — they all end up at the same place:
use Siberfx\Typesense\Standalone\TypesenseManager;
// 1. Facade — proxies to the default connection
TypesenseDirect::search('books', ['q' => 'dune', 'query_by' => 'title']);
// 2. A specific named connection
TypesenseDirect::connection('analytics')->search('events', ['q' => '*']);
// 3. Container / dependency injection
$ts = app('typesense.manager'); // or: app(TypesenseManager::class)
$ts->connection()->listCollections();$connection = TypesenseDirect::connection(); // default connection
// Create a collection by hand
$connection->createCollection([
'name' => 'books',
'fields' => [
['name' => 'title', 'type' => 'string'],
['name' => 'author', 'type' => 'string', 'facet' => true],
['name' => 'year', 'type' => 'int32', 'sort' => true],
],
'default_sorting_field' => 'year',
]);
// Create only if missing (retrieve-or-create)
$connection->ensureCollection([ 'name' => 'books', 'fields' => [/* … */] ]);
$connection->hasCollection('books'); // bool
$connection->retrieveCollection('books'); // schema + stats
$connection->listCollections(); // all collections
// Add a field (alter)
$connection->alterCollection('books', [
'fields' => [['name' => 'in_stock', 'type' => 'bool']],
]);
$connection->dropCollection('books'); // delete the collectiondocuments(string $collection) returns a small helper bound to one collection:
$docs = TypesenseDirect::documents('books');
$docs->create(['id' => '1', 'title' => 'Dune', 'year' => 1965]);
$docs->upsert(['id' => '1', 'title' => 'Dune', 'year' => 1965]); // create or replace
$docs->update(['id' => '1', 'year' => 1966]); // partial update
$docs->retrieve('1'); // single document
$docs->delete('1'); // by id
// Bulk import — array of docs OR a JSONL string.
// $action: 'create' | 'upsert' | 'update' | 'emplace'
$results = $docs->import([
['id' => '1', 'title' => 'Dune', 'year' => 1965],
['id' => '2', 'title' => 'Neuromancer', 'year' => 1984],
], 'upsert');
// Delete many by filter
$docs->deleteByFilter(['filter_by' => 'year:<1950']);
// Export the whole collection as a JSONL string
$jsonl = $docs->export();// Single search
$hits = TypesenseDirect::search('books', [
'q' => 'dune',
'query_by' => 'title',
'filter_by' => 'year:>1900',
'sort_by' => 'year:desc',
'per_page' => 20,
]);
echo $hits['found']; // hit count
$hits['hits'][0]['document']; // the matched document
// Federated multi-search — pass the list of searches; the second arg holds
// parameters common to all of them. Results come back under $res['results'].
$res = TypesenseDirect::multiSearch(
[
['collection' => 'books', 'q' => 'dune', 'query_by' => 'title'],
['collection' => 'authors', 'q' => 'herbert','query_by' => 'name'],
],
['per_page' => 5] // common params
);
$res['results'][0]['hits'];Generate a scoped key that embeds search parameters (e.g. a tenant filter_by
and/or an expires_at). Computed locally via HMAC — no API call:
$scoped = TypesenseDirect::generateScopedSearchKey($parentSearchKey, [
'filter_by' => 'company_id:42',
'expires_at' => now()->addDay()->timestamp,
]);These accessors return the native typesense-php resource objects, so the
full underlying API is available:
$c = TypesenseDirect::connection();
// API keys
$c->keys()->create(['description' => 'search-only', 'actions' => ['documents:search'], 'collections' => ['*']]);
// Collection aliases
$c->aliases()->upsert('books', ['collection_name' => 'books_v2']);
// Search presets
$c->presets()->upsert('popular', ['value' => ['query_by' => 'title', 'sort_by' => '_text_match:desc']]);
// Stopwords (note: this resource uses put/get/getAll/delete)
$c->stopwords()->put(['name' => 'stw_en', 'stopwords' => ['a', 'the'], 'locale' => 'en']);
// Stemming dictionaries
$c->stemming()->dictionaries()->upsert('irregulars', [['word' => 'people', 'root' => 'person']]);
// Analytics rules
$c->analytics()->rules()->upsert('popular_queries', ['type' => 'popular_queries', 'params' => [/* … */]]);
// Conversation (RAG) & natural-language search models
$c->conversations()->getModels()->retrieve();
$c->nlSearchModels()->retrieve();
// Cluster ops
$c->health()->retrieve();
$c->metrics()->retrieve();
$c->operations();
$c->debug();For anything not wrapped (e.g. per-collection synonyms/overrides), reach the
underlying \Typesense\Client directly:
$client = TypesenseDirect::connection()->client();
$client->getCollections()['books']->getSynonyms()->upsert('coat-synonyms', [
'synonyms' => ['blazer', 'coat', 'jacket'],
]);| Area | Methods on TypesenseDirect::connection() |
|---|---|
| Collections | createCollection, ensureCollection, hasCollection, retrieveCollection, alterCollection, dropCollection, listCollections |
Documents (documents($c)->) |
create, upsert, update, retrieve, delete, deleteByFilter, import, export |
| Search | search, multiSearch |
| Keys | keys, generateScopedSearchKey |
| Admin | aliases, presets, stopwords, stemming, analytics, analyticsV1, synonymSets, curationSets, conversations, nlSearchModels |
| Ops | health, metrics, debug, operations |
| Escape hatch | client() |
Note
This standalone client targets typesense/typesense-php ^6 (server v30+).
That unlocks the global resources below; on older servers, pin the client to
^5 and use the client() escape hatch instead.
Typesense v30 promotes synonyms and curation to shareable, top-level resources, reachable directly from the connection:
$c = TypesenseDirect::connection();
// Global synonym set (link to a collection via its `synonym_sets` field)
$c->synonymSets()->upsert('clothing', [
'items' => [
['id' => 'coats', 'synonyms' => ['blazer', 'coat', 'jacket']],
],
]);
// Global curation set (link via a collection's `curation_sets` field)
$c->curationSets()->upsert('promos', [
'items' => [[
'id' => 'apple-promo',
'rule' => ['query' => 'apple', 'match' => 'exact'],
'includes' => [['id' => '422', 'position' => 1]],
'excludes' => [['id' => '287']],
]],
]);
// Legacy analytics surface (alongside analytics())
$c->analyticsV1()->rules()->retrieve();- Replace
siberfx/laravel-typesensein your composer.json requirements withsiberfx/typesense-scout - The Scout driver is now called
typesense, instead oftypesensesearch. This should be reflected by setting the SCOUT_DRIVER env var totypesense, and changing the config/scout.php config key fromtypesensesearchtotypesense - Instead of importing
Siberfx\Typesense\*, you should importSiberfx\Typesense\* - Instead of models implementing
Siberfx\Typesense\Interfaces\TypesenseSearch, they should implementSiberfx\Typesense\Interfaces\TypesenseDocument
composer install
vendor/bin/phpunitThe suite has two groups:
-
Unit — pure logic (filter building, scoped key generation, config shape, public API surface). No server required.
-
Integration — end-to-end flows against a real Typesense server (collections, documents, filtered search, synonyms, overrides, aliases, presets, stopwords). These are skipped automatically when no server is reachable. Point them at a server via environment variables:
TYPESENSE_HOST=localhost TYPESENSE_PORT=8108 \ TYPESENSE_PROTOCOL=http TYPESENSE_API_KEY=xyz \ vendor/bin/phpunit --testsuite Integration
Anonymous
Other key contributors include:
The MIT License (MIT). Please see the License File for more information.