Algolia Search is a hosted full-text, numerical, and faceted search engine capable of delivering realtime results from the first keystroke.
This gem let you easily integrate the Algolia Search API to your favorite ORM. It's based on the algoliasearch-client-ruby gem. Rails 3.x, 4.x and 5.x are all supported.
You might be interested in the sample Ruby on Rails application providing a autocomplete.js
-based auto-completion and instantsearch.js
-based instant search results page: algoliasearch-rails-example.
You can find the full reference on Algolia's website.
gem install algoliasearch-rails
Add the gem to your Gemfile
:
gem "algoliasearch-rails"
And run:
bundle install
Create a new file config/initializers/algoliasearch.rb
to setup your APPLICATION_ID
and API_KEY
.
AlgoliaSearch.configuration = { application_id: 'YourApplicationID', api_key: 'YourAPIKey' }
The gem is compatible with ActiveRecord, Mongoid and Sequel.
You can configure a various timeout thresholds by setting the following options at initialization time:
AlgoliaSearch.configuration = {
application_id: 'YourApplicationID',
api_key: 'YourAPIKey',
connect_timeout: 2,
receive_timeout: 30,
send_timeout: 30,
batch_timeout: 120,
search_timeout: 5
}
This gem makes extensive use of Rails' callbacks to trigger the indexing tasks. If you're using methods bypassing after_validation
, before_save
or after_commit
callbacks, it will not index your changes. For example: update_attribute
doesn't perform validations checks, to perform validations when updating use update_attributes
.
All methods injected by the AlgoliaSearch
module are prefixed by algolia_
and aliased to the associated short names if they aren't already defined.
Contact.algolia_reindex! # <=> Contact.reindex!
Contact.algolia_search("jon doe") # <=> Contact.search("jon doe")
The following code will create a Contact
index and add search capabilities to your Contact
model:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
attribute :first_name, :last_name, :email
end
end
You can either specify the attributes to send (here we restricted to :first_name, :last_name, :email
) or not (in that case, all attributes are sent).
class Product < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
# all attributes will be sent
end
end
You can also use the add_attribute
method, to send all model attributes + extra ones:
class Product < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
# all attributes + extra_attr will be sent
add_attribute :extra_attr
end
def extra_attr
"extra_val"
end
end
We provide many ways to configure your index allowing you to tune your overall index relevancy. The most important ones are the searchable attributes and the attributes reflecting record popularity.
class Product < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
# list of attribute used to build an Algolia record
attributes :title, :subtitle, :description, :likes_count, :seller_name
# the `searchableAttributes` (formerly known as attributesToIndex) setting defines the attributes
# you want to search in: here `title`, `subtitle` & `description`.
# You need to list them by order of importance. `description` is tagged as
# `unordered` to avoid taking the position of a match into account in that attribute.
searchableAttributes ['title', 'subtitle', 'unordered(description)']
# the `customRanking` setting defines the ranking criteria use to compare two matching
# records in case their text-relevance is equal. It should reflect your record popularity.
customRanking ['desc(likes_count)']
end
end
To index a model, simple call reindex
on the class:
Product.reindex
To index all of your models, you can do something like this:
Rails.application.eager_load! # Ensure all models are loaded (required in development).
algolia_models = ActiveRecord::Base.descendants.select{ |model| model.respond_to?(:reindex) }
algolia_models.each(&:reindex)
Traditional search implementations tend to have search logic and functionality on the backend. This made sense when the search experience consisted of a user entering a search query, executing that search, and then being redirected to a search result page.
Implementing search on the backend is no longer necessary. In fact, in most cases it is harmful to performance because of added network and processing latency. We highly recommend the usage of our JavaScript API Client issuing all search requests directly from the end user's browser, mobile device, or client. It will reduce the overall search latency while offloading your servers at the same time.
The JS API client is part of the gem, just require algolia/v3/algoliasearch.min
somewhere in your JavaScript manifest, for example in application.js
if you are using Rails 3.1+:
//= require algolia/v3/algoliasearch.min
Then in your JavaScript code you can do:
var client = algoliasearch(ApplicationID, Search-Only-API-Key);
var index = client.initIndex('YourIndexName');
index.search('something', { hitsPerPage: 10, page: 0 })
.then(function searchDone(content) {
console.log(content)
})
.catch(function searchFailure(err) {
console.error(err);
});
We recently (March 2015) released a new version (V3) of our JavaScript client, if you were using our previous version (V2), read the migration guide
Notes: We recommend the usage of our JavaScript API Client to perform queries directly from the end-user browser without going through your server.
A search returns ORM-compliant objects reloading them from your database. We recommend the usage of our JavaScript API Client to perform queries to decrease the overall latency and offload your servers.
hits = Contact.search("jon doe")
p hits
p hits.raw_answer # to get the original JSON raw answer
A highlight_result
attribute is added to each ORM object:
hits[0].highlight_result['first_name']['value']
If you want to retrieve the raw JSON answer from the API, without re-loading the objects from the database, you can use:
json_answer = Contact.raw_search("jon doe")
p json_answer
p json_answer['hits']
p json_answer['facets']
Search parameters can be specified either through the index's settings statically in your model or dynamically at search time specifying search parameters as second argument of the search
method:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
attribute :first_name, :last_name, :email
# default search parameters stored in the index settings
minWordSizefor1Typo 4
minWordSizefor2Typos 8
hitsPerPage 42
end
end
# dynamical search parameters
p Contact.raw_search('jon doe', { hitsPerPage: 5, page: 2 })
Even if we highly recommend to perform all search (and therefore pagination) operations from your frontend using JavaScript, we support both will_paginate and kaminari as pagination backend.
To use :will_paginate
, specify the :pagination_backend
as follow:
AlgoliaSearch.configuration = { application_id: 'YourApplicationID', api_key: 'YourAPIKey', pagination_backend: :will_paginate }
Then, as soon as you use the search
method, the returning results will be a paginated set:
# in your controller
@results = MyModel.search('foo', hitsPerPage: 10)
# in your views
# if using will_paginate
<%= will_paginate @results %>
# if using kaminari
<%= paginate @results %>
Use the tags
method to add tags to your record:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
tags ['trusted']
end
end
or using dynamical values:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
tags do
[first_name.blank? || last_name.blank? ? 'partial' : 'full', has_valid_email? ? 'valid_email' : 'invalid_email']
end
end
end
At query time, specify { tagFilters: 'tagvalue' }
or { tagFilters: ['tagvalue1', 'tagvalue2'] }
as search parameters to restrict the result set to specific tags.
Facets can be retrieved calling the extra facets
method of the search answer.
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
# [...]
# specify the list of attributes available for faceting
attributesForFaceting [:company, :zip_code]
end
end
hits = Contact.search('jon doe', { facets: '*' })
p hits # ORM-compliant array of objects
p hits.facets # extra method added to retrieve facets
p hits.facets['company'] # facet values+count of facet 'company'
p hits.facets['zip_code'] # facet values+count of facet 'zip_code'
raw_json = Contact.raw_search('jon doe', { facets: '*' })
p raw_json['facets']
You can also search for facet values.
Product.search_for_facet_values('category', 'Headphones') # Array of {value, highlighted, count}
This method can also take any parameter a query can take. This will adjust the search to only hits which would have matched the query.
# Only sends back the categories containing red Apple products (and only counts those)
Product.search_for_facet_values('category', 'phone', {
query: 'red',
filters: 'brand:Apple'
}) # Array of phone categories linked to red Apple products
More info on distinct for grouping can be found here.
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
# [...]
# specify the attribute to be used for distinguishing the records
# in this case the records will be grouped by company
attributeForDistinct "company"
end
end
Use the geoloc
method to localize your record:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
geoloc :lat_attr, :lng_attr
end
end
At query time, specify { aroundLatLng: "37.33, -121.89", aroundRadius: 50000 }
as search parameters to restrict the result set to 50KM around San Jose.
Each time a record is saved; it will be - asynchronously - indexed. On the other hand, each time a record is destroyed, it will be - asynchronously - removed from the index. That means that a network call with the ADD/DELETE operation is sent synchronously to the Algolia API but then the engine will asynchronously process the operation (so if you do a search just after, the results may not reflect it yet).
You can disable auto-indexing and auto-removing setting the following options:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch auto_index: false, auto_remove: false do
attribute :first_name, :last_name, :email
end
end
You can temporary disable auto-indexing using the without_auto_index
scope. This is often used for performance reason.
Contact.delete_all
Contact.without_auto_index do
1.upto(10000) { Contact.create! attributes } # inside this block, auto indexing task will not run.
end
Contact.reindex! # will use batch operations
You can configure the auto-indexing & auto-removal process to use a queue to perform those operations in background. ActiveJob (Rails >=4.2) queues are used by default but you can define your own queuing mechanism:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch enqueue: true do # ActiveJob will be triggered using a `algoliasearch` queue
attribute :first_name, :last_name, :email
end
end
If you are performing updates & deletions in the background then a record deletion can be committed to your database prior to the job actually executing. Thus if you were to load the record to remove it from the database than your ActiveRecord#find will fail with a RecordNotFound.
In this case you can bypass loading the record from ActiveRecord and just communicate with the index directly:
class MySidekiqWorker
def perform(id, remove)
if remove
# the record has likely already been removed from your database so we cannot
# use ActiveRecord#find to load it
index = Algolia::Index.new("index_name")
index.delete_object(id)
else
# the record should be present
c = Contact.find(id)
c.index!
end
end
end
If you're using Sidekiq:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch enqueue: :trigger_sidekiq_worker do
attribute :first_name, :last_name, :email
end
def self.trigger_sidekiq_worker(record, remove)
MySidekiqWorker.perform_async(record.id, remove)
end
end
class MySidekiqWorker
def perform(id, remove)
if remove
# the record has likely already been removed from your database so we cannot
# use ActiveRecord#find to load it
index = Algolia::Index.new("index_name")
index.delete_object(id)
else
# the record should be present
c = Contact.find(id)
c.index!
end
end
end
If you're using delayed_job:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch enqueue: :trigger_delayed_job do
attribute :first_name, :last_name, :email
end
def self.trigger_delayed_job(record, remove)
if remove
record.delay.remove_from_index!
else
record.delay.index!
end
end
end
You can force indexing and removing to be synchronous (in that case the gem will call the wait_task
method to ensure the operation has been taken into account once the method returns) by setting the following option: (this is NOT recommended, except for testing purpose)
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch synchronous: true do
attribute :first_name, :last_name, :email
end
end
By default, the index name will be the class name, e.g. "Contact". You can customize the index name by using the index_name
option:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch index_name: "MyCustomName" do
attribute :first_name, :last_name, :email
end
end
You can suffix the index name with the current Rails environment using the following option:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch per_environment: true do # index name will be "Contact_#{Rails.env}"
attribute :first_name, :last_name, :email
end
end
You can use a block to specify a complex attribute value
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
attribute :email
attribute :full_name do
"#{first_name} #{last_name}"
end
add_attribute :full_name2
end
def full_name2
"#{first_name} #{last_name}"
end
end
Notes: As soon as you use such code to define extra attributes, the gem is not anymore able to detect if the attribute has changed (the code uses Rails's #{attribute}_changed?
method to detect that). As a consequence, your record will be pushed to the API even if its attributes didn't change. You can work-around this behavior creating a _changed?
method:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch do
attribute :email
attribute :full_name do
"#{first_name} #{last_name}"
end
end
def full_name_changed?
first_name_changed? || last_name_changed?
end
end
You can easily embed nested objects defining an extra attribute returning any JSON-compliant object (an array or a hash or a combination of both).
class Profile < ActiveRecord::Base
include AlgoliaSearch
belongs_to :user
has_many :specializations
algoliasearch do
attribute :user do
# restrict the nested "user" object to its `name` + `email`
{ name: user.name, email: user.email }
end
attribute :public_specializations do
# build an array of public specialization (include only `title` and `another_attr`)
specializations.select { |s| s.public? }.map do |s|
{ title: s.title, another_attr: s.another_attr }
end
end
end
end
By default, the objectID
is based on your record's id
. You can change this behavior specifying the :id
option (be sure to use a uniq field).
class UniqUser < ActiveRecord::Base
include AlgoliaSearch
algoliasearch id: :uniq_name do
end
end
You can add constraints controlling if a record must be indexed by using options the :if
or :unless
options.
It allows you to do conditional indexing on a per document basis.
class Post < ActiveRecord::Base
include AlgoliaSearch
algoliasearch if: :published?, unless: :deleted? do
end
def published?
# [...]
end
def deleted?
# [...]
end
end
Notes: As soon as you use those constraints, addObjects
and deleteObjects
calls will be performed in order to keep the index synced with the DB (The state-less gem doesn't know if the object don't match your constraints anymore or never matched, so we force ADD/DELETE operations to be sent). You can work-around this behavior creating a _changed?
method:
class Contact < ActiveRecord::Base
include AlgoliaSearch
algoliasearch if: :published do
end
def published
# true or false
end
def published_changed?
# return true only if you know that the 'published' state changed
end
end
You can index a subset of your records using either:
# will generate batch API calls (recommended)
MyModel.where('updated_at > ?', 10.minutes.ago).reindex!
or
MyModel.index_objects MyModel.limit(5)
You can sanitize all your attributes using the sanitize
option. It will strip all HTML tags from your attributes.
class User < ActiveRecord::Base
include AlgoliaSearch
algoliasearch per_environment: true, sanitize: true do
attributes :name, :email, :company
end
end
If you're using Rails 4.2+, you also need to depend on rails-html-sanitizer
:
gem 'rails-html-sanitizer'
You can force the UTF-8 encoding of all your attributes using the force_utf8_encoding
option:
class User < ActiveRecord::Base
include AlgoliaSearch
algoliasearch force_utf8_encoding: true do
attributes :name, :email, :company
end
end
Notes: This option is not compatible with Ruby 1.8
You can disable exceptions that could be raised while trying to reach Algolia's API by using the raise_on_failure
option:
class Contact < ActiveRecord::Base
include AlgoliaSearch
# only raise exceptions in development env
algoliasearch raise_on_failure: Rails.env.development? do
attribute :first_name, :last_name, :email
end
end
Here is a real-word configuration example (from HN Search):
class Item < ActiveRecord::Base
include AlgoliaSearch
algoliasearch per_environment: true do
# the list of attributes sent to Algolia's API
attribute :created_at, :title, :url, :author, :points, :story_text, :comment_text, :author, :num_comments, :story_id, :story_title
# integer version of the created_at datetime field, to use numerical filtering
attribute :created_at_i do
created_at.to_i
end
# `title` is more important than `{story,comment}_text`, `{story,comment}_text` more than `url`, `url` more than `author`
# btw, do not take into account position in most fields to avoid first word match boost
searchableAttributes ['unordered(title)', 'unordered(story_text)', 'unordered(comment_text)', 'unordered(url)', 'author']
# tags used for filtering
tags do
[item_type, "author_#{author}", "story_#{story_id}"]
end
# use associated number of HN points to sort results (last sort criteria)
customRanking ['desc(points)', 'desc(num_comments)']
# google+, $1.5M raises, C#: we love you
separatorsToIndex '+#$'
end
def story_text
item_type_cd != Item.comment ? text : nil
end
def story_title
comment? && story ? story.title : nil
end
def story_url
comment? && story ? story.url : nil
end
def comment_text
comment? ? text : nil
end
def comment?
item_type_cd == Item.comment
end
# [...]
end
You can trigger indexing using the index!
instance method.
c = Contact.create!(params[:contact])
c.index!
And trigger index removing using the remove_from_index!
instance method.
c.remove_from_index!
c.destroy
The gem provides 2 ways to reindex all your objects:
To reindex all your records (taking into account the deleted objects), the reindex
class method indices all your objects to a temporary index called <INDEX_NAME>.tmp
and moves the temporary index to the final one once everything is indexed (atomically). This is the safest way to reindex all your content.
Contact.reindex
Notes: if you're using an index-specific API key, ensure you're allowing both <INDEX_NAME>
and <INDEX_NAME>.tmp
.
Warning: You should not use such an atomic reindexing operation while scoping/filtering the model because this operation replaces the entire index, keeping the filtered objects only. ie: Don't do MyModel.where(...).reindex
but do MyModel.where(...).reindex!
(with the trailing !
)!!!
To reindex all your objects in place (without temporary index and therefore without deleting removed objects), use the reindex!
class method:
Contact.reindex!
To clear an index, use the clear_index!
class method:
Contact.clear_index!
You can access the underlying index
object by calling the index
class method:
index = Contact.index
# index.get_settings, index.partial_update_object, ...
You can define replica indices using the add_replica
method.
Use inherit: true
on the replica block if you want it to inherit from the primary settings.
class Book < ActiveRecord::Base
attr_protected
include AlgoliaSearch
algoliasearch per_environment: true do
searchableAttributes [:name, :author, :editor]
# define a replica index to search by `author` only
add_replica 'Book_by_author', per_environment: true do
searchableAttributes [:author]
end
# define a replica index with custom ordering but same settings than the main block
add_replica 'Book_custom_order', inherit: true, per_environment: true do
customRanking ['asc(rank)']
end
end
end
To search using a replica, use the following code:
Book.raw_search 'foo bar', replica: 'Book_by_editor'
# or
Book.search 'foo bar', replica: 'Book_by_editor'
It can make sense to share an index between several models. In order to implement that, you'll need to ensure you don't have any conflict with the objectID
of the underlying models.
class Student < ActiveRecord::Base
attr_protected
include AlgoliaSearch
algoliasearch index_name: 'people', id: :algolia_id do
# [...]
end
private
def algolia_id
"student_#{id}" # ensure the teacher & student IDs are not conflicting
end
end
class Teacher < ActiveRecord::Base
attr_protected
include AlgoliaSearch
algoliasearch index_name: 'people', id: :algolia_id do
# [...]
end
private
def algolia_id
"teacher_#{id}" # ensure the teacher & student IDs are not conflicting
end
end
Notes: If you target a single index from several models, you must never use MyModel.reindex
and only use MyModel.reindex!
. The reindex
method uses a temporary index to perform an atomic reindexing: if you use it, the resulting index will only contain records for the current model because it will not reindex the others.
You can index a record in several indices using the add_index
method:
class Book < ActiveRecord::Base
attr_protected
include AlgoliaSearch
PUBLIC_INDEX_NAME = "Book_#{Rails.env}"
SECURED_INDEX_NAME = "SecuredBook_#{Rails.env}"
# store all books in index 'SECURED_INDEX_NAME'
algoliasearch index_name: SECURED_INDEX_NAME do
searchableAttributes [:name, :author]
# convert security to tags
tags do
[released ? 'public' : 'private', premium ? 'premium' : 'standard']
end
# store all 'public' (released and not premium) books in index 'PUBLIC_INDEX_NAME'
add_index PUBLIC_INDEX_NAME, if: :public? do
searchableAttributes [:name, :author]
end
end
private
def public?
released && !premium
end
end
To search using an extra index, use the following code:
Book.raw_search 'foo bar', index: 'Book_by_editor'
# or
Book.search 'foo bar', index: 'Book_by_editor'
To run the specs, please set the ALGOLIA_APPLICATION_ID
and ALGOLIA_API_KEY
environment variables. Since the tests are creating and removing indices, DO NOT use your production account.
You may want to disable all indexing (add, update & delete operations) API calls, you can set the disable_indexing
option:
class User < ActiveRecord::Base
include AlgoliaSearch
algoliasearch per_environment: true, disable_indexing: Rails.env.test? do
end
end
class User < ActiveRecord::Base
include AlgoliaSearch
algoliasearch per_environment: true, disable_indexing: Proc.new { Rails.env.test? || more_complex_condition } do
end
end
Or you may want to mock Algolia's API calls. We provide a WebMock sample configuration that you can use including algolia/webmock
:
require 'algolia/webmock'
describe 'With a mocked client' do
before(:each) do
WebMock.enable!
end
it "shouldn't perform any API calls here" do
User.create(name: 'My Indexed User') # mocked, no API call performed
User.search('').should == {} # mocked, no API call performed
end
after(:each) do
WebMock.disable!
end
end