Module: LlmCostTracker::Integrations::RubyLlm

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

Defined Under Namespace

Modules: ProviderPatch

Constant Summary

Constants included from Base

Base::Result

Class Method Summary collapse

Methods included from Base

active?, enforce_budget!, install, minimum_version, normalize_sdk_args, object_dig, object_value, patch_target, patch_targets, record_safely, request_params, status, stream_collector, stream_pricing_mode, track_stream, version_constant

Class Method Details

.image_token_detail(usage, direction) ⇒ Object 



101
102
103
104
105
106
107
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 101

def image_token_detail(usage, direction)
  container_key = direction == :input ? :input_tokens_details : :output_tokens_details
  details = usage[container_key] || usage[container_key.to_s] || {}
  return 0 unless details.is_a?(Hash)

  (details[:image_tokens] || details["image_tokens"]).to_i
end

.integration_name ⇒ Object 



12
13
14
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 12

def integration_name
  :ruby_llm
end

.minimum_version ⇒ Object 



16
17
18
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 16

def minimum_version
  "1.14.1"
end

.model_id(object) ⇒ Object 



172
173
174
175
176
177
178
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 172

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 



24
25
26
27
28
29
30
31
32
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 24

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

.pricing_mode(provider:, response:) ⇒ Object 



189
190
191
192
193
194
195
196
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 189

def pricing_mode(provider:, response:)
  raw = object_value(response, :pricing_mode, :service_tier)
  if provider == "anthropic" && LlmCostTracker::Providers::Anthropic::TierClassification.standard_equivalent_tier?(raw)
    return nil
  end

  raw
end

.provider_response_id(response) ⇒ Object 



185
186
187
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 185

def provider_response_id(response)
  object_value(response, :id, :provider_response_id)
end

.provider_slug(provider) ⇒ Object 



168
169
170
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 168

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

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



34
35
36
37
38
39
40
41
42
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 34

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 



48
49
50
51
52
53
54
55
56
57
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 48

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_image(provider, response, request:, latency_ms:) ⇒ Object 



69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 69

def record_image(provider, response, request:, latency_ms:)
  usage = object_value(response, :usage)
  usage = {} unless usage.is_a?(Hash)
  raw_input = (usage[:input_tokens] || usage["input_tokens"]).to_i
  raw_output = (usage[:output_tokens] || usage["output_tokens"]).to_i
  image_input = image_token_detail(usage, :input)
  image_output = image_token_detail(usage, :output)
  text_input = [raw_input - image_input, 0].max
  text_output = [raw_output - image_output, 0].max
  record_passthrough(
    provider: provider_slug(provider),
    model: response_model_id(response) || model_id(request[:model]),
    response: response,
    latency_ms: latency_ms,
    input_tokens: text_input,
    image_input_tokens: image_input,
    output_tokens: text_output,
    image_output_tokens: image_output
  )
end

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



90
91
92
93
94
95
96
97
98
99
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 90

def record_moderation(provider, response, request:, latency_ms:)
  record_passthrough(
    provider: provider_slug(provider),
    model: response_model_id(response) || model_id(request[:model]),
    response: response,
    latency_ms: latency_ms,
    input_tokens: 0,
    output_tokens: 0
  )
end

.record_passthrough(provider:, model:, response:, latency_ms:, input_tokens:, output_tokens:, image_input_tokens: 0, image_output_tokens: 0) ⇒ Object 



109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 109

def record_passthrough(provider:, model:, response:, latency_ms:, input_tokens:, output_tokens:,
                       image_input_tokens: 0, image_output_tokens: 0)
  return unless active?

  record_safely do
    LlmCostTracker::Tracker.record(
      event: Event.build(
        provider: provider,
        model: model,
        token_usage: TokenUsage.build(
          input_tokens: input_tokens,
          output_tokens: output_tokens,
          image_input_tokens: image_input_tokens,
          image_output_tokens: image_output_tokens
        ),
        usage_source: :sdk_response,
        provider_response_id: provider_response_id(response)
      ),
      latency_ms: latency_ms
    )
  end
end

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



59
60
61
62
63
64
65
66
67
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 59

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 



132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 132

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(
      event: Event.build(
        provider: provider,
        model: model,
        pricing_mode: pricing_mode(provider: provider, response: 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: :sdk_response,
        provider_response_id: provider_response_id(response)
      ),
      latency_ms: latency_ms
    )
  end
end

.regular_input_tokens(input_tokens, cache_read) ⇒ Object 



164
165
166
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 164

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

.response_model_id(object) ⇒ Object 



180
181
182
183
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 180

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

.streaming_request?(request, has_block:) ⇒ Boolean

Returns:

  • (Boolean) 


44
45
46
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 44

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

.version_constant ⇒ Object 



20
21
22
 
# File 'lib/llm_cost_tracker/integrations/ruby_llm.rb', line 20

def version_constant
  "RubyLLM::VERSION"
end