Setting up the OpenAI client

Add this line to your application's Gemfile:

  gem "ruby-openai"

If you would like to override the default OpenAI client configuration options, you can set them in an initializer .

# config/initializers/openai.rb - example rails initializer
OpenAI.configure do |config|
  config.access_token = Rails.application.credentials.openai[:access_token]

  # As of 2024-12-13 you can set the following client configuration options:
  # config.api_type
  # config.api_version
  # config.access_token
  # config.organization_id
  # config.uri_base
  # config.request_timeout
  # config.extra_headers
end

Alternatively, instruct will use an access token set in OPENAI_ACCESS_TOKEN or OPENAI_API_KEY environment variables.

Using completions API instead of chat

The ruby-openai gem allows calls to be made to the completions endpoint instead of the chat api. Currently gpt-3.5-turbo-instruct is the only model that uses this API and OpenAI has marked it as deprecated, neverthless, if the gem is used with a different base_uri, such as with a local model, you can force use of the completions endpoint by setting the use_completion_endpoint: true keyword argument when creating a model with Instruct::OpenAI#new or in a #gen call that resolves to an Instruct::OpenAI model.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!