llm_cost_tracker 0.15.1 → 0.15.3
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +4 -4
- data/CHANGELOG.md +24 -0
- data/lib/llm_cost_tracker/integrations/openai/patches.rb +6 -0
- data/lib/llm_cost_tracker/integrations/openai.rb +6 -0
- data/lib/llm_cost_tracker/integrations/ruby_llm/v2/attempt/token_counts.rb +2 -1
- data/lib/llm_cost_tracker/integrations/ruby_llm/v2/attempt.rb +15 -5
- data/lib/llm_cost_tracker/integrations/ruby_llm/v2/frame.rb +11 -0
- data/lib/llm_cost_tracker/integrations/ruby_llm/v2/seams.rb +17 -1
- data/lib/llm_cost_tracker/prices.json +698 -248
- data/lib/llm_cost_tracker/providers/openai/parser.rb +1 -1
- data/lib/llm_cost_tracker/providers/openai/response_parser.rb +11 -1
- data/lib/llm_cost_tracker/version.rb +1 -1
- metadata +1 -1
|
@@ -8,7 +8,7 @@ module LlmCostTracker
|
|
|
8
8
|
include StreamParser
|
|
9
9
|
|
|
10
10
|
TRACKED_ENDPOINTS = %w[
|
|
11
|
-
chat/completions completions embeddings moderations responses responses/compact
|
|
11
|
+
chat/completions completions decisions embeddings moderations responses responses/compact
|
|
12
12
|
audio/transcriptions audio/translations audio/speech
|
|
13
13
|
images/generations images/edits images/variations
|
|
14
14
|
].freeze
|
|
@@ -14,6 +14,7 @@ module LlmCostTracker
|
|
|
14
14
|
PENDING_STATUSES = %w[queued in_progress].freeze
|
|
15
15
|
RETRIEVE_PATH_PATTERN = %r{/(?:responses|agent)/resp_}
|
|
16
16
|
TRANSCRIPTION_PATH_PATTERN = %r{/audio/(?:transcriptions|translations)\z}
|
|
17
|
+
UNBILLED_DECISION_USAGE = %i[output_tokens output_tokens_details total_tokens].freeze
|
|
17
18
|
|
|
18
19
|
class << self
|
|
19
20
|
def included(base) = base.include(StreamParser)
|
|
@@ -29,7 +30,7 @@ module LlmCostTracker
|
|
|
29
30
|
end
|
|
30
31
|
|
|
31
32
|
def event_from_response(response:, request:, provider:, host:, usage_source:, pricing_mode: nil)
|
|
32
|
-
usage = response
|
|
33
|
+
usage = billed_usage(response, provider)
|
|
33
34
|
return nil if usage.nil? || pending?(response)
|
|
34
35
|
|
|
35
36
|
model = response["model"] || request["model"]
|
|
@@ -73,6 +74,15 @@ module LlmCostTracker
|
|
|
73
74
|
def pending?(response)
|
|
74
75
|
PENDING_STATUSES.include?(response["status"].to_s)
|
|
75
76
|
end
|
|
77
|
+
|
|
78
|
+
def billed_usage(response, provider)
|
|
79
|
+
usage = response["usage"]&.deep_symbolize_keys
|
|
80
|
+
return usage unless usage && decision?(response, provider)
|
|
81
|
+
|
|
82
|
+
usage.except(*UNBILLED_DECISION_USAGE)
|
|
83
|
+
end
|
|
84
|
+
|
|
85
|
+
def decision?(response, provider) = provider.to_s == "openai" && response.key?("answers")
|
|
76
86
|
end
|
|
77
87
|
|
|
78
88
|
def parse(request_url:, request_body:, response_status:, response_body:, **)
|