Module: LlmCostTracker::Integrations::RubyLlm

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

Defined Under Namespace

Modules: ProviderPatch

Class Method Summary collapse

Methods included from Base

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

Class Method Details

.integration_name ⇒ Object 



11
12
13
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 11

def integration_name
  :ruby_llm
end

.minimum_version ⇒ Object 



15
16
17
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 15

def minimum_version
  "1.14.1"
end

.model_id(object) ⇒ Object 



108
109
110
111
112
113
114
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 108

def model_id(object)
  return nil if object.nil?

  value = object_value(object, :id, :model_id, :model)
  value ||= object if object.is_a?(String) || object.is_a?(Symbol)
  value&.to_s
end

.patch_targets ⇒ Object 



23
24
25
26
27
28
29
30
31
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 23

def patch_targets
  [
    patch_target(
      "RubyLLM::Provider",
      with: ProviderPatch,
      methods: %i[slug complete embed transcribe]
    )
  ]
end

.pricing_mode(response) ⇒ Object 



125
126
127
128
129
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 125

def pricing_mode(response)
  object_value(response, :pricing_mode, :service_tier) ||
    object_dig(response, :raw, :pricing_mode) ||
    object_dig(response, :raw, :service_tier)
end

.provider_response_id(response) ⇒ Object 



121
122
123
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 121

def provider_response_id(response)
  object_value(response, :id, :provider_response_id) || object_dig(response, :raw, :id)
end

.provider_slug(provider) ⇒ Object 



104
105
106
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 104

def provider_slug(provider)
  object_value(provider, :slug).to_s
end

.record_completion(provider, response, request:, latency_ms:, stream:) ⇒ Object 



33
34
35
36
37
38
39
40
41
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 33

def record_completion(provider, response, request:, latency_ms:, stream:)
  record_usage(
    provider: provider_slug(provider),
    model: response_model_id(response) || model_id(request[:model]),
    response: response,
    latency_ms: latency_ms,
    stream: stream
  )
end

.record_embedding(provider, response, request:, latency_ms:) ⇒ Object 



47
48
49
50
51
52
53
54
55
56
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 47

def record_embedding(provider, response, request:, latency_ms:)
  record_usage(
    provider: provider_slug(provider),
    model: response_model_id(response) || model_id(request[:model]),
    response: response,
    latency_ms: latency_ms,
    stream: false,
    output_tokens: 0
  )
end

.record_transcription(provider, response, request:, latency_ms:) ⇒ Object 



58
59
60
61
62
63
64
65
66
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 58

def record_transcription(provider, response, request:, latency_ms:)
  record_usage(
    provider: provider_slug(provider),
    model: response_model_id(response) || model_id(request[:model]),
    response: response,
    latency_ms: latency_ms,
    stream: false
  )
end

.record_usage(provider:, model:, response:, latency_ms:, stream:, output_tokens: nil) ⇒ Object 



68
69
70
71
72
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
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 68

def record_usage(provider:, model:, response:, latency_ms:, stream:, output_tokens: nil)
  return unless active?

  record_safely do
    input_tokens = object_value(response, :input_tokens)
    output_tokens = object_value(response, :output_tokens) if output_tokens.nil?
    next if input_tokens.nil? && output_tokens.nil?

    cache_read = object_value(response, :cached_tokens).to_i
    hidden_output = object_value(response, :thinking_tokens, :reasoning_tokens).to_i

    LlmCostTracker::Tracker.record(
      capture: UsageCapture.build(
        provider: provider,
        model: model,
        pricing_mode: pricing_mode(response),
        token_usage: TokenUsage.build(
          input_tokens: regular_input_tokens(input_tokens, cache_read),
          output_tokens: output_tokens.to_i,
          cache_read_input_tokens: cache_read,
          cache_write_input_tokens: object_value(response, :cache_creation_tokens).to_i,
          hidden_output_tokens: hidden_output
        ),
        stream: stream,
        usage_source: :ruby_llm,
        provider_response_id: provider_response_id(response)
      ),
      latency_ms: latency_ms
    )
  end
end

.regular_input_tokens(input_tokens, cache_read) ⇒ Object 



100
101
102
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 100

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

.response_model_id(object) ⇒ Object 



116
117
118
119
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 116

def response_model_id(object)
  value = object_value(object, :model_id, :model)
  value&.to_s
end

.streaming_request?(request, has_block:) ⇒ Boolean

Returns:

  • (Boolean) 


43
44
45
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 43

def streaming_request?(request, has_block:)
  has_block || request[:stream] == true
end

.version_constant ⇒ Object 



19
20
21
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 19

def version_constant
  "RubyLLM::VERSION"
end