idea1.rb

require "bundler/inline"

gemfile do
  source "https://rubygems.org"

  gem "activerecord", require: "active_record"
  gem "neighbor", github: "ankane/neighbor"
  gem "pg"
end

ActiveRecord::Base.establish_connection adapter: "postgresql", database: "neighbor_example"
ActiveRecord::Base.logger = ActiveSupport::Logger.new(STDOUT)

ActiveRecord::Schema.define do
  enable_extension "vector"

  create_table :todos, force: :cascade do |t|
    t.column :embedding, :vector, limit: 3
  end
end

class Todo < ActiveRecord::Base
  has_neighbors :embedding

  default_scope { select(column_names - ["embedding"]) }
end

def generate_embedding
  3.times.map { rand }
end

3.times do
  Todo.create!(embedding: generate_embedding)
end

# loads embedding for nearest neighbor searches only
todo = Todo.except(:select).find(1)
pp todo.nearest_neighbors(:embedding, distance: "cosine").first(5)

# does not load embedding (Rails 7.1+)
todo = Todo.find(1)
pp Todo.with(neighbor_cte: Todo.reselect(:embedding).where(id: todo.id))
    .where.not(id: todo.id)
    .where.not(embedding: nil)
    .order(Arel.sql("embedding <=> (SELECT embedding FROM neighbor_cte)"))
    .first(5)

idea2.rb

require "bundler/inline"

gemfile do
  source "https://rubygems.org"

  gem "activerecord", require: "active_record"
  gem "neighbor"
  gem "pg"
end

ActiveRecord::Base.establish_connection adapter: "postgresql", database: "neighbor_example"
ActiveRecord::Base.logger = ActiveSupport::Logger.new(STDOUT)

ActiveRecord::Schema.define do
  enable_extension "vector"

  create_table :todos, force: :cascade do |t|
  end

  create_table :todo_embeddings, force: :cascade do |t|
    t.belongs_to :todo
    t.column :embedding, :vector, limit: 3
  end
end

class Todo < ActiveRecord::Base
  has_one :todo_embedding
end

class TodoEmbedding < ActiveRecord::Base
  belongs_to :todo

  has_neighbors :embedding
end

def generate_embedding
  3.times.map { rand }
end

3.times do
  t = Todo.create!
  t.create_todo_embedding!(embedding: generate_embedding)
end

todo = Todo.find(1)

# loads embedding for nearest neighbor searches only
pp todo
    .todo_embedding
    .nearest_neighbors(:embedding, distance: "cosine")
    .reselect(:todo_id)
    .includes(:todo)
    .first(5)
    .map(&:todo)

# does not load embedding (Rails 7.1+)
pp Todo.with(neighbor_cte: TodoEmbedding.select(:embedding).where(todo_id: todo.id))
    .joins(:todo_embedding)
    .where.not(id: todo.id)
    .where.not(todo_embedding: {embedding: nil})
    .order(Arel.sql("embedding <=> (SELECT embedding FROM neighbor_cte)"))
    .first(5)