NameError: uninitialized class variable @@provider

[andreibondarev]

I ran into an issue when attempting to run 'SupportArticle.embed!' via the Rails console.

I get the following error:

SupportArticle Load (9.3ms)  SELECT "support_articles".* FROM "support_articles" ORDER BY "support_articles"."id" ASC LIMIT $1  [["LIMIT", 1000]]
NameError: uninitialized class variable @@provider in SupportArticle

from /Users/williamahlering/.gem/ruby/3.2.2/gems/langchainrb_rails-0.1.10/lib/langchainrb_rails/active_record/hooks.rb:52:in `class_variable_get'

I followed the setup instructions to add pgvector and successfully migrated the db to add the 'embedding' vector column to my model 'SupportArticle'. I've also set my ENV["OPENAI_API_KEY"] which I can see correctly in the rails console, as well.

My gemfile includes the following:

gem 'pgvector'
gem 'neighbor'
gem "langchainrb_rails", "~> 0.1.10"
gem "ruby-openai", "~> 7.1"
gem 'ollama-ai', '~> 1.3.0'

[sergiobayona]

just some notes of minor issues that need to address but can be fixed manually:

doing:

rails generate langchainrb_rails:pgvector --model=SupportArticle --llm=openai

didn't underscore the table name:

class AddVectorColumnToSupportarticles < ActiveRecord::Migration[7.1]
  def change
    add_column :supportarticles, :embedding, :vector,
      limit: LangchainrbRails
        .config
        .vectorsearch
        .llm
        .default_dimensions
  end
end

didn't add the openai gem either.

[rickbenavidez]

If anyone else runs into the @@Provider exception message that @andreibondarev mentions here, double check that you've also explicitly added vectorsearch to the corresponding model file where your embedding column has been added. This error occurs when that piece is missing and you attempt to run any of the embeddings commands against it. I'm not sure if this helps with the original issue, but wanted to share where I saw it while getting things set up.

[andreibondarev]

If anyone else runs into the @@Provider exception message that @andreibondarev mentions here, double check that you've also explicitly added vectorsearch to the corresponding model file where your embedding column has been added. This error occurs when that piece is missing and you attempt to run any of the embeddings commands against it. I'm not sure if this helps with the original issue, but wanted to share where I saw it while getting things set up.

I'm wondering if there's a way to gracefully handle this error, like rescue-ing and then raising a specific error that points out that your model file is missing the vectorsearch line.

[emk-klaay]

If I run the agent from within a long chain tool, even though the inventor macro call is there, the @@provider is still nil.