MONGOID-5923 Support auto emdedding vector search indexes and searches

[jamis]

Summary

Mongoid can now define vector search indexes with auto-embedding, and provides a helper interface for querying documents by their semantic meaning using those indexed vector fields.

The simplest interface will define a String field using the voyage-4 model.

class Article
  include Mongoid::Document
  
  # Define `body` as a String field, indexed using the `voyage-4` auto-embedding model.
  auto_embed_field :body
end

You may declare the model and other parameters as well, if desired:

class Article
  include Mongoid::Document

  auto_embed_field :body,
                       model: 'voyage-4-large'
                       num_dimensions: 512,
                       quantization: 'binary',
                       similarity: 'cosine',
                       index: :body_index
end

Then, create the new search indexes:

$ rake db:mongoid:create_search_indexes

Once the indexes are created, you can query them using the auto_embed_search method:

articles = Article.auto_embed_search('machine learning', limit: 5)

# or, search for documents that are similar to an existing document
neighbors = article.auto_embed_search(limit: 5)

[Copilot]

Pull request overview

Adds Mongoid support for Atlas Vector Search auto-embedding indexes (autoEmbed) and a convenience query API to search by semantic text similarity.

Changes:

• Introduces auto_embed_field to declare a String field and register a vectorSearch index spec with an autoEmbed field definition.
• Adds .auto_embed_search and #auto_embed_search APIs to execute $vectorSearch auto-embedding queries (including filtering, exact mode, and optional model override).
• Expands RSpec coverage with argument-validation tests and (currently skipped) Atlas integration test scaffolding for auto-embed indexes.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
lib/mongoid/fields.rb	Adds auto_embed_field macro that defines a text field and registers an autoEmbed vector search index spec.
lib/mongoid/search_indexable.rb	Adds class/instance auto_embed_search implementations plus index/path inference helpers for auto-embed specs.
spec/mongoid/fields_spec.rb	Adds unit specs asserting index spec registration, defaults, and field behavior for auto_embed_field.
spec/mongoid/search_indexable_spec.rb	Adds argument validation specs for auto_embed_search and scaffolds skipped Atlas integration coverage.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.