Module: LlmCostTracker::Integrations::Openai

Extended by:
Base
Defined in:
lib/llm_cost_tracker/integrations/openai.rb

Defined Under Namespace

Modules: ChatCompletionsPatch, ResponsesPatch

Class Method Summary collapse

Methods included from Base

active?, constant, elapsed_ms, enforce_budget!, install, minimum_version, patch_target, patch_targets, record_safely, request_params, status, version_constant

Class Method Details

.cache_read_input_tokens(usage) ⇒ Object 



56
57
58
59
60
61
 
# File 'lib/llm_cost_tracker/integrations/openai.rb', line 56

def cache_read_input_tokens(usage)
  ObjectReader.integer(
    ObjectReader.nested(usage, :input_tokens_details, :cached_tokens) ||
    ObjectReader.nested(usage, :prompt_tokens_details, :cached_tokens)
  )
end

.hidden_output_tokens(usage) ⇒ Object 



63
64
65
66
67
68
 
# File 'lib/llm_cost_tracker/integrations/openai.rb', line 63

def hidden_output_tokens(usage)
  ObjectReader.integer(
    ObjectReader.nested(usage, :output_tokens_details, :reasoning_tokens) ||
    ObjectReader.nested(usage, :completion_tokens_details, :reasoning_tokens)
  )
end

.integration_name ⇒ Object 



11
 
# File 'lib/llm_cost_tracker/integrations/openai.rb', line 11

def integration_name = :openai

.minimum_version ⇒ Object 



13
 
# File 'lib/llm_cost_tracker/integrations/openai.rb', line 13

def minimum_version = "0.59.0"

.patch_targets ⇒ Object 



17
18
19
20
21
22
 
# File 'lib/llm_cost_tracker/integrations/openai.rb', line 17

def patch_targets
  [
    patch_target("OpenAI::Resources::Responses", with: ResponsesPatch, methods: :create),
    patch_target("OpenAI::Resources::Chat::Completions", with: ChatCompletionsPatch, methods: :create)
  ]
end

.record_response(response, request:, latency_ms:) ⇒ Object 



24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
 
# File 'lib/llm_cost_tracker/integrations/openai.rb', line 24

def record_response(response, request:, latency_ms:)
  return unless active?

  record_safely do
    usage = ObjectReader.first(response, :usage)
    next unless usage

    input_tokens = ObjectReader.first(usage, :input_tokens, :prompt_tokens)
    output_tokens = ObjectReader.first(usage, :output_tokens, :completion_tokens)
    next if input_tokens.nil? && output_tokens.nil?

    metadata = usage_metadata(usage)
    LlmCostTracker::Tracker.record(
      provider: "openai",
      model: ObjectReader.first(response, :model) || request[:model],
      input_tokens: regular_input_tokens(input_tokens, metadata[:cache_read_input_tokens]),
      output_tokens: ObjectReader.integer(output_tokens),
      latency_ms: latency_ms,
      usage_source: :sdk_response,
      provider_response_id: ObjectReader.first(response, :id),
      metadata: metadata
    )
  end
end

.regular_input_tokens(input_tokens, cache_read) ⇒ Object 



70
71
72
 
# File 'lib/llm_cost_tracker/integrations/openai.rb', line 70

def regular_input_tokens(input_tokens, cache_read)
  [ObjectReader.integer(input_tokens) - cache_read.to_i, 0].max
end

.usage_metadata(usage) ⇒ Object 



49
50
51
52
53
54
 
# File 'lib/llm_cost_tracker/integrations/openai.rb', line 49

def usage_metadata(usage)
  {
    cache_read_input_tokens: cache_read_input_tokens(usage),
    hidden_output_tokens: hidden_output_tokens(usage)
  }
end

.version_constant ⇒ Object 



15
 
# File 'lib/llm_cost_tracker/integrations/openai.rb', line 15

def version_constant = "OpenAI::VERSION"