Class: OpenaiEmbeddingCreator

Inherits:
Object
  • Object
show all
Defined in:
lib/embedding_engines/openai_embeddings.rb

Instance Method Summary collapse

Constructor Details

#initialize(api_key = nil, chunker = BasicTextChunker.new, model = "text-embedding-ada-002") ⇒ OpenaiEmbeddingCreator

Returns a new instance of OpenaiEmbeddingCreator.



8
9
10
11
12
13
14
15
16
17
18
19
 
# File 'lib/embedding_engines/openai_embeddings.rb', line 8

def initialize(api_key = nil, chunker = BasicTextChunker.new, model = "text-embedding-ada-002")
  @chunker = chunker
  @model = model
  @api_key = api_key || ENV['OPENAI_API_KEY']
  raise 'API key not found. Please set the OPENAI_API_KEY environment variable.' if api_key.nil? || api_key.empty?

  if @api_key
    @llm = OpenAI::Client.new(access_token: @api_key)
  else
    Rails.logger.error "OpenAI API key not provided. Set the OPENAI_API_KEY in the ENV variables or pass it as an argument."
  end
end

Instance Method Details

#call(text, pages_mode = true) ⇒ Object 



21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
 
# File 'lib/embedding_engines/openai_embeddings.rb', line 21

def call(text, pages_mode=true)

  if pages_mode
    vectors = []
    return [] unless @llm  # Return empty if the API client isn't set up

    # Divide the text into chunks for each page
    text.each_with_index do |page_content, page_index|
      chunks = @chunker.split_into_chunks(page_content)

      # Create embeddings for each chunk
      chunks.each_with_index do |chunk, index|
        response = @llm.embeddings(
          parameters: {
            model: @model,
            input: chunk
          }
        )

        # Extract the embeddings from the response
        embedding = response['data'][0]['embedding']

        # Create vector data for the chunk and keep page numbers for reference
        vector_data = {
          id: "vec #{index + 1}",
          values: embedding,
          metadata: {
              text: chunk,
              page: page_index + 1,
          }
        }
        # storing each chunk vector data in an array
        vectors << vector_data
      end
    end
    vectors
  else
    response = @llm.embeddings(
      parameters: {
        model: @model,
        input: chunk
      }
    )
    # Extract the embeddings from the response
    response['data'][0]['embedding']
  end
end