Class: Ragdoll::EmbeddingService

Inherits:
Object
  • Object
show all
Defined in:
app/services/ragdoll/embedding_service.rb

Instance Method Summary collapse

Constructor Details

#initialize(client: nil, config_service: nil, model_resolver: nil) ⇒ EmbeddingService

Returns a new instance of EmbeddingService.



7
8
9
10
11
12
 
# File 'app/services/ragdoll/embedding_service.rb', line 7

def initialize(client: nil, config_service: nil, model_resolver: nil)
  @client = client
  @config_service = config_service || Ragdoll::ConfigurationService.new
  @model_resolver = model_resolver || Ragdoll::ModelResolver.new(@config_service)
  configure_ruby_llm unless @client
end

Instance Method Details

#cosine_similarity(embedding1, embedding2) ⇒ Object 



133
134
135
136
137
138
139
140
141
142
143
144
 
# File 'app/services/ragdoll/embedding_service.rb', line 133

def cosine_similarity(embedding1, embedding2)
  return 0.0 if embedding1.nil? || embedding2.nil?
  return 0.0 if embedding1.length != embedding2.length

  dot_product = embedding1.zip(embedding2).sum { |a, b| a * b }
  magnitude1 = Math.sqrt(embedding1.sum { |a| a * a })
  magnitude2 = Math.sqrt(embedding2.sum { |a| a * a })

  return 0.0 if magnitude1 == 0.0 || magnitude2 == 0.0

  dot_product / (magnitude1 * magnitude2)
end

#generate_embedding(text) ⇒ Object 



14
15
16
17
18
19
20
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
68
69
70
71
 
# File 'app/services/ragdoll/embedding_service.rb', line 14

def generate_embedding(text)
  return nil if text.nil? || text.strip.empty?

  # Clean and prepare text
  cleaned_text = clean_text(text)

  begin
    if @client
      # Use custom client for testing
      embedding_config = @model_resolver.resolve_embedding(:text)
      response = @client.embed(
        input: cleaned_text,
        model: embedding_config.model.to_s
      )

      if response && response["embeddings"]&.first
        response["embeddings"].first
      elsif response && response["data"]&.first && response["data"].first["embedding"]
        response["data"].first["embedding"]
      else
        raise Ragdoll::Core::EmbeddingError, "Invalid response format from embedding API"
      end
    else
      # Use RubyLLM for real embedding generation
      embedding_config = @model_resolver.resolve_embedding(:text)
      # Use just the model name for RubyLLM
      model = embedding_config.model.model
      
      # If model is nil or empty, use fallback
      if model.nil? || model.empty?
        return generate_fallback_embedding
      end

      begin
        response = RubyLLM.embed(cleaned_text, model: model)

        # Extract the embedding vector from RubyLLM::Embedding object
        return generate_fallback_embedding unless response.respond_to?(:instance_variable_get)

        vectors = response.instance_variable_get(:@vectors)
        return generate_fallback_embedding unless vectors && vectors.is_a?(Array)

        vectors
      rescue StandardError
        # If RubyLLM fails, use fallback
        generate_fallback_embedding
      end
    end
  rescue StandardError => e
    # Only use fallback if no client was provided (RubyLLM failures)
    # If a client was provided, we should raise the error for proper test behavior
    raise Ragdoll::Core::EmbeddingError, "Failed to generate embedding: #{e.message}" if @client

    # No client - this is a RubyLLM configuration issue, use fallback
    puts "Warning: Embedding generation failed (#{e.message}), using fallback"
    generate_fallback_embedding
  end
end

#generate_embeddings_batch(texts) ⇒ Object 



73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
 
# File 'app/services/ragdoll/embedding_service.rb', line 73

def generate_embeddings_batch(texts)
  return [] if texts.empty?

  # Clean all texts
  cleaned_texts = texts.map { |text| clean_text(text) }.reject { |t| t.nil? || t.strip.empty? }
  return [] if cleaned_texts.empty?

  begin
    if @client
      # Use custom client for testing
      embedding_config = @model_resolver.resolve_embedding(:text)
      response = @client.embed(
        input: cleaned_texts,
        model: embedding_config.model.to_s
      )

      if response && response["embeddings"]
        response["embeddings"]
      elsif response && response["data"]
        response["data"].map { |item| item["embedding"] }
      else
        raise Ragdoll::Core::EmbeddingError, "Invalid response format from embedding API"
      end
    else
      # Use RubyLLM for real embedding generation (batch mode)
      embedding_config = @model_resolver.resolve_embedding(:text)
      # Use just the model name for RubyLLM
      model = embedding_config.model.model
      
      # If model is nil or empty, use fallback
      if model.nil? || model.empty?
        return cleaned_texts.map { generate_fallback_embedding }
      end

      cleaned_texts.map do |text|
        response = RubyLLM.embed(text, model: model)

        # Extract the embedding vector from RubyLLM::Embedding object
        next generate_fallback_embedding unless response.respond_to?(:instance_variable_get)

        vectors = response.instance_variable_get(:@vectors)
        next generate_fallback_embedding unless vectors && vectors.is_a?(Array)

        vectors
      rescue StandardError
        # If RubyLLM fails, use fallback
        generate_fallback_embedding
      end
    end
  rescue StandardError => e
    # Only use fallback if no client was provided (RubyLLM failures)
    # If a client was provided, we should raise the error for proper test behavior
    raise Ragdoll::Core::EmbeddingError, "Failed to generate embeddings: #{e.message}" if @client

    # No client - this is a RubyLLM configuration issue, use fallback
    puts "Warning: Batch embedding generation failed (#{e.message}), using fallback"
    texts.map { generate_fallback_embedding }
  end
end