chatwoot-develop/enterprise/app/services/captain/llm/embedding_service.rb

54 lines
1.7 KiB
Ruby
Executable File

class Captain::Llm::EmbeddingService
include Integrations::LlmInstrumentation
class EmbeddingsError < StandardError; end
def initialize(account_id: nil)
Llm::Config.initialize!
@account_id = account_id
@embedding_model = InstallationConfig.find_by(name: 'CAPTAIN_EMBEDDING_MODEL')&.value.presence || LlmConstants::DEFAULT_EMBEDDING_MODEL
end
def self.embedding_model
InstallationConfig.find_by(name: 'CAPTAIN_EMBEDDING_MODEL')&.value.presence || LlmConstants::DEFAULT_EMBEDDING_MODEL
end
def get_embedding(content, model: @embedding_model)
return generate_fallback_embedding('empty') if content.blank?
instrument_embedding_call(instrumentation_params(content, model)) do
response = RubyLLM.embed(content, model: model)
return response.vectors.flatten if response.vectors.present? && response.vectors.first.present?
Rails.logger.warn 'OpenAI returned empty embedding, using fallback'
generate_fallback_embedding(content)
end
rescue StandardError => e
Rails.logger.error "Embedding API/DB Error: #{e.message}, using fallback"
generate_fallback_embedding(content)
end
private
def generate_fallback_embedding(text)
# Deterministic fallback for stability
require 'digest'
seed = Digest::SHA256.hexdigest(text.to_s.downcase.strip).to_i(16) % (2**32)
rng = Random.new(seed)
# OpenAI default dimensions
vector = Array.new(1536) { rng.rand(-1.0..1.0) }
magnitude = Math.sqrt(vector.sum { |v| v**2 })
vector.map { |v| v / magnitude }
end
def instrumentation_params(content, model)
{
span_name: 'llm.captain.embedding',
model: model,
input: content,
feature_name: 'embedding',
account_id: @account_id
}
end
end