llm_cost_tracker 0.4.1 → 0.5.1

data/lib/llm_cost_tracker/tag_context.rb

@@ -0,0 +1,52 @@
+# frozen_string_literal: true
+
+require "active_support/isolated_execution_state"
+
+require_relative "value_helpers"
+
+module LlmCostTracker
+  module TagContext
+    KEY = :llm_cost_tracker_tags
+
+    class << self
+      def with(tags)
+        stack = current_stack
+        ActiveSupport::IsolatedExecutionState[KEY] = stack + [normalize(tags)]
+        yield
+      ensure
+        ActiveSupport::IsolatedExecutionState[KEY] = stack
+      end
+
+      def tags
+        config_tags.merge(scoped_tags)
+      end
+
+      def clear!
+        ActiveSupport::IsolatedExecutionState[KEY] = []
+      end
+
+      private
+
+      def config_tags
+        normalize(resolve_default_tags)
+      end
+
+      def resolve_default_tags
+        tags = LlmCostTracker.configuration.default_tags
+        tags.respond_to?(:call) ? tags.call : tags
+      end
+
+      def scoped_tags
+        current_stack.reduce({}) { |merged, tags| merged.merge(tags) }
+      end
+
+      def current_stack
+        ActiveSupport::IsolatedExecutionState[KEY] || []
+      end
+
+      def normalize(tags)
+        ValueHelpers.deep_dup(tags || {}).to_h
+      end
+    end
+  end
+end

data/lib/llm_cost_tracker/tracker.rb

@@ -1,12 +1,12 @@
 # frozen_string_literal: true
 
-require_relative "logging"
+require_relative "storage/dispatcher"
 
 module LlmCostTracker
   class Tracker
     EVENT_NAME = "llm_request.llm_cost_tracker"
 
-    USAGE_SOURCES = %i[response stream_final manual unknown].freeze
+    USAGE_SOURCES = %i[response stream_final sdk_response manual unknown].freeze
 
     class << self
       def enforce_budget!
@@ -19,6 +19,7 @@ module LlmCostTracker
                  usage_source: nil, provider_response_id: nil, pricing_mode: nil, metadata: {})
         return unless LlmCostTracker.configuration.enabled
 
+        model = normalize_model(model)
         usage = usage_data(input_tokens, output_tokens, metadata, pricing_mode)
         cost_data = cost_for_usage(provider, model, usage)
 
@@ -38,7 +39,7 @@ module LlmCostTracker
 
         ActiveSupport::Notifications.instrument(EVENT_NAME, event.to_h)
 
-        stored = store(event)
+        stored = Storage::Dispatcher.save(event)
         Budget.check!(event) unless stored == false
 
         event
@@ -68,6 +69,8 @@ module LlmCostTracker
         )
       end
 
+      def normalize_model(value) = value.to_s.strip.then { |model| model.empty? ? ParsedUsage::UNKNOWN_MODEL : model }
+
       def build_event(provider:, model:, usage:, cost_data:, metadata:, latency_ms:, stream:, usage_source:,
                       provider_response_id:)
         Event.new(
@@ -81,7 +84,7 @@ module LlmCostTracker
           hidden_output_tokens: usage[:hidden_output_tokens],
           pricing_mode: usage[:pricing_mode],
           cost: cost_data,
-          tags: LlmCostTracker.configuration.default_tags.merge(EventMetadata.tags(metadata)).freeze,
+          tags: LlmCostTracker::TagContext.tags.merge(EventMetadata.tags(metadata)).freeze,
           latency_ms: normalized_latency_ms(latency_ms),
           stream: stream ? true : false,
           usage_source: normalized_usage_source(usage_source),
@@ -90,62 +93,6 @@ module LlmCostTracker
         )
       end
 
-      def store(event)
-        config = LlmCostTracker.configuration
-        case config.storage_backend
-        when :log            then log_event(event, config)
-        when :active_record  then active_record_save(event)
-        when :custom         then custom_save(event, config)
-        end
-      rescue BudgetExceededError, UnknownPricingError
-        raise
-      rescue StandardError => e
-        handle_storage_error(e)
-        false
-      end
-
-      def log_event(event, config)
-        message = "#{event.provider}/#{event.model} " \
-                  "tokens=#{event.total_tokens} " \
-                  "cost=#{log_cost_label(event)}"
-        message += " latency=#{event.latency_ms}ms" if event.latency_ms
-        message += " stream=#{event.stream}" if event.stream
-        message += " source=#{event.usage_source}" if event.usage_source
-        message += " tags=#{event.tags}" unless event.tags.empty?
-
-        Logging.log(config.log_level, message)
-        event
-      end
-
-      def log_cost_label(event) = event.cost ? "$#{format('%.6f', event.cost.total_cost)}" : "unknown"
-
-      def active_record_save(event)
-        require_relative "llm_api_call" unless defined?(LlmCostTracker::LlmApiCall)
-        require_relative "storage/active_record_store" unless defined?(LlmCostTracker::Storage::ActiveRecordStore)
-
-        Storage::ActiveRecordStore.save(event)
-        event
-      rescue LoadError => e
-        raise Error, "ActiveRecord storage requires the active_record gem: #{e.message}"
-      end
-
-      def custom_save(event, config)
-        result = config.custom_storage&.call(event)
-        result == false ? false : event
-      end
-
-      def handle_storage_error(error)
-        case LlmCostTracker.configuration.storage_error_behavior
-        when :ignore
-          nil
-        when :warn
-          Logging.warn("Storage failed; tracking event was not persisted: #{error.class}: #{error.message}")
-        when :raise
-          storage_error = StorageError.new(error)
-          raise storage_error
-        end
-      end
-
       def normalized_latency_ms(latency_ms) = latency_ms.nil? ? nil : [latency_ms.to_i, 0].max
 
       def normalized_usage_source(value)

data/lib/llm_cost_tracker/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module LlmCostTracker
-  VERSION = "0.4.1"
+  VERSION = "0.5.1"
 end

data/lib/llm_cost_tracker.rb

@@ -25,9 +25,11 @@ require_relative "llm_cost_tracker/parsers/gemini"
 require_relative "llm_cost_tracker/parsers/sse"
 require_relative "llm_cost_tracker/parsers/registry"
 require_relative "llm_cost_tracker/middleware/faraday"
+require_relative "llm_cost_tracker/integrations/registry"
 require_relative "llm_cost_tracker/budget"
 require_relative "llm_cost_tracker/unknown_pricing"
 require_relative "llm_cost_tracker/event_metadata"
+require_relative "llm_cost_tracker/tag_context"
 require_relative "llm_cost_tracker/tags_column"
 require_relative "llm_cost_tracker/tag_key"
 require_relative "llm_cost_tracker/tag_query"
@@ -37,6 +39,7 @@ require_relative "llm_cost_tracker/retention"
 require_relative "llm_cost_tracker/report_data"
 require_relative "llm_cost_tracker/report_formatter"
 require_relative "llm_cost_tracker/report"
+require_relative "llm_cost_tracker/doctor"
 
 module LlmCostTracker
   CONFIGURATION_MUTEX = Monitor.new
@@ -52,10 +55,11 @@ module LlmCostTracker
         current = current.dup_for_configuration if current.finalized?
         @configuration = current
         yield(current)
-        current.normalize_openai_compatible_providers!
+        current.openai_compatible_providers = current.openai_compatible_providers.dup
         current.finalize!
         current
       end
+      Integrations.install!
       warn_for_configuration!(config)
     end
 
@@ -63,14 +67,20 @@ module LlmCostTracker
       CONFIGURATION_MUTEX.synchronize { @configuration = Configuration.new }
       UnknownPricing.reset! if defined?(UnknownPricing)
       Storage::ActiveRecordStore.reset! if defined?(Storage::ActiveRecordStore)
+      TagContext.clear! if defined?(TagContext)
     end
 
     def enforce_budget!
       Tracker.enforce_budget!
     end
 
-    def track(provider:, model:, input_tokens:, output_tokens:, latency_ms: nil, stream: false, usage_source: :manual,
-              enforce_budget: false, provider_response_id: nil, pricing_mode: nil, **metadata)
+    def with_tags(tags = nil, **kwargs, &)
+      merged = (tags || {}).to_h.merge(kwargs)
+      TagContext.with(merged, &)
+    end
+
+    def track(provider:, input_tokens:, output_tokens:, model: nil, latency_ms: nil, stream: false,
+              usage_source: :manual, enforce_budget: false, provider_response_id: nil, pricing_mode: nil, **metadata)
       enforce_budget! if enforce_budget
       Tracker.record(
         provider: provider.to_s,
@@ -86,7 +96,7 @@ module LlmCostTracker
       )
     end
 
-    def track_stream(provider:, model:, latency_ms: nil, enforce_budget: false, provider_response_id: nil,
+    def track_stream(provider:, model: nil, latency_ms: nil, enforce_budget: false, provider_response_id: nil,
                      pricing_mode: nil, **metadata)
       require_relative "llm_cost_tracker/stream_collector"
       enforce_budget! if enforce_budget

data/lib/tasks/llm_cost_tracker.rake

@@ -1,7 +1,17 @@
 # frozen_string_literal: true
 
+require "fileutils"
+
 # rubocop:disable Metrics/BlockLength
 namespace :llm_cost_tracker do
+  desc "Check LLM Cost Tracker setup"
+  task :doctor do
+    Rake::Task["environment"].invoke if Rake::Task.task_defined?("environment")
+    checks = LlmCostTracker::Doctor.call
+    puts LlmCostTracker::Doctor.report(checks)
+    abort("llm_cost_tracker: doctor found setup errors") unless LlmCostTracker::Doctor.healthy?(checks)
+  end
+
   desc "Print an LLM cost report from ActiveRecord storage"
   task report: :environment do
     days = (ENV["DAYS"] || LlmCostTracker::Report::DEFAULT_DAYS).to_i
@@ -18,56 +28,49 @@ namespace :llm_cost_tracker do
 
   namespace :prices do
     desc(
-      "Sync built-in pricing data from LiteLLM/OpenRouter JSON sources. " \
-      "Use PREVIEW=1 to preview, STRICT=1 to fail on provider errors, " \
-      "or OUTPUT=path/to/file.json."
+      "Refresh the configured pricing file from the maintained LLM Cost Tracker price snapshot. " \
+      "Use PREVIEW=1 to preview, URL=... to override the source, or OUTPUT=path/to/file.json."
     )
-    task :sync do
+    task :refresh do
       Rake::Task["environment"].invoke if Rake::Task.task_defined?("environment")
       require_relative "../llm_cost_tracker"
 
-      output_path = ENV["OUTPUT"] || LlmCostTracker.configuration.prices_file || LlmCostTracker::PriceSync::DEFAULT_OUTPUT_PATH
-      strict = ENV["STRICT"] == "1" || ARGV.include?("--strict")
-      result = LlmCostTracker::PriceSync.sync(
+      output_path = price_refresh_output_path
+      source_url = LlmCostTracker::PriceSync.configured_remote_url
+      preview = ENV["PREVIEW"] == "1"
+      result = LlmCostTracker::PriceSync.refresh(
         path: output_path,
-        preview: ENV["PREVIEW"] == "1",
-        strict: strict
+        url: source_url,
+        preview: preview
       )
 
-      action = if ENV["PREVIEW"] == "1"
+      action = if preview
                  "previewed"
                elsif result.written
-                 "updated"
+                 "refreshed"
                else
                  "kept"
                end
 
       puts "llm_cost_tracker: #{action} pricing file #{result.path}"
-      print_source_usage(result.sources_used)
+      puts "  source: #{result.source_url}"
+      puts "  version: #{result.source_version.inspect}" if result.source_version
       print_changes(result.changes)
-      print_discrepancies(result.discrepancies)
-      print_issues("validator rejected", result.rejected)
-      print_issues("validator flagged", result.flagged)
-      print_models("orphaned models (no JSON source match)", result.orphaned_models)
-      print_failures(result.failed_sources, heading: "source failures (kept existing values)")
     end
 
-    desc "Compare the current pricing snapshot with LiteLLM/OpenRouter JSON sources and exit non-zero on drift."
+    desc "Compare the current pricing file with the maintained LLM Cost Tracker price snapshot."
     task :check do
       Rake::Task["environment"].invoke if Rake::Task.task_defined?("environment")
       require_relative "../llm_cost_tracker"
 
-      output_path = ENV["OUTPUT"] || LlmCostTracker.configuration.prices_file || LlmCostTracker::PriceSync::DEFAULT_OUTPUT_PATH
-      result = LlmCostTracker::PriceSync.check(path: output_path)
+      output_path = price_refresh_output_path
+      source_url = LlmCostTracker::PriceSync.configured_remote_url
+      result = LlmCostTracker::PriceSync.check(path: output_path, url: source_url)
 
       puts "llm_cost_tracker: checked pricing file #{result.path}"
-      print_source_usage(result.sources_used)
+      puts "  source: #{result.source_url}"
+      puts "  version: #{result.source_version.inspect}" if result.source_version
       print_changes(result.changes)
-      print_discrepancies(result.discrepancies)
-      print_issues("validator rejected", result.rejected)
-      print_issues("validator flagged", result.flagged)
-      print_models("orphaned models (no JSON source match)", result.orphaned_models)
-      print_failures(result.failed_sources, heading: "source failures")
       puts "  pricing is up to date" if result.up_to_date
       abort("llm_cost_tracker: pricing check failed") unless result.up_to_date
     end
@@ -75,16 +78,6 @@ namespace :llm_cost_tracker do
 end
 # rubocop:enable Metrics/BlockLength
 
-def print_source_usage(sources_used)
-  return if sources_used.empty?
-
-  puts "  sources used:"
-  sources_used.each do |source, usage|
-    version = usage.source_version ? ", version=#{usage.source_version.inspect}" : ""
-    puts "    - #{source} (#{usage.prices_count} prices#{version})"
-  end
-end
-
 def print_changes(changes)
   puts "  changed models: #{changes.size}"
   return if changes.empty?
@@ -97,37 +90,8 @@ def print_changes(changes)
   end
 end
 
-def print_discrepancies(discrepancies)
-  return if discrepancies.empty?
-
-  puts "  source discrepancies: #{discrepancies.size}"
-  discrepancies.each do |issue|
-    formatted = issue.values.map { |source, value| "#{source}=#{value.inspect}" }.join(", ")
-    puts "    - #{issue.model} #{issue.field}: #{formatted}"
-  end
-end
-
-def print_issues(heading, issues)
-  return if issues.empty?
-
-  puts "  #{heading}: #{issues.size}"
-  issues.each do |issue|
-    puts "    - #{issue.model}: #{issue.reason}"
-  end
-end
-
-def print_models(heading, models)
-  return if models.empty?
-
-  puts "  #{heading}: #{models.size}"
-  models.each { |model| puts "    - #{model}" }
-end
-
-def print_failures(failed_sources, heading:)
-  return if failed_sources.empty?
-
-  puts "  #{heading}: #{failed_sources.size}"
-  failed_sources.each do |source, message|
-    puts "    - #{source}: #{message}"
-  end
+def price_refresh_output_path
+  path = LlmCostTracker::PriceSync.configured_output_path
+  FileUtils.mkdir_p(File.dirname(path))
+  path
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: llm_cost_tracker
 version: !ruby/object:Gem::Version
-  version: 0.4.1
+  version: 0.5.1
 platform: ruby
 authors:
 - Sergii Khomenko
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-04-24 00:00:00.000000000 Z
+date: 2026-04-27 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: activesupport
@@ -84,6 +84,20 @@ dependencies:
     - - "<"
       - !ruby/object:Gem::Version
         version: '9.0'
+- !ruby/object:Gem::Dependency
+  name: nokogiri
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '1.16'
+  type: :development
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '1.16'
 - !ruby/object:Gem::Dependency
   name: railties
   requirement: !ruby/object:Gem::Requirement
@@ -272,7 +286,9 @@ files:
 - lib/llm_cost_tracker/assets.rb
 - lib/llm_cost_tracker/budget.rb
 - lib/llm_cost_tracker/configuration.rb
+- lib/llm_cost_tracker/configuration/instrumentation.rb
 - lib/llm_cost_tracker/cost.rb
+- lib/llm_cost_tracker/doctor.rb
 - lib/llm_cost_tracker/engine.rb
 - lib/llm_cost_tracker/engine_compatibility.rb
 - lib/llm_cost_tracker/errors.rb
@@ -292,11 +308,15 @@ files:
 - lib/llm_cost_tracker/generators/llm_cost_tracker/templates/add_usage_breakdown_to_llm_api_calls.rb.erb
 - lib/llm_cost_tracker/generators/llm_cost_tracker/templates/create_llm_api_calls.rb.erb
 - lib/llm_cost_tracker/generators/llm_cost_tracker/templates/initializer.rb.erb
-- lib/llm_cost_tracker/generators/llm_cost_tracker/templates/llm_cost_tracker_prices.yml.erb
 - lib/llm_cost_tracker/generators/llm_cost_tracker/templates/upgrade_llm_api_call_cost_precision.rb.erb
 - lib/llm_cost_tracker/generators/llm_cost_tracker/templates/upgrade_llm_api_call_tags_to_jsonb.rb.erb
 - lib/llm_cost_tracker/generators/llm_cost_tracker/upgrade_cost_precision_generator.rb
 - lib/llm_cost_tracker/generators/llm_cost_tracker/upgrade_tags_to_jsonb_generator.rb
+- lib/llm_cost_tracker/integrations/anthropic.rb
+- lib/llm_cost_tracker/integrations/base.rb
+- lib/llm_cost_tracker/integrations/object_reader.rb
+- lib/llm_cost_tracker/integrations/openai.rb
+- lib/llm_cost_tracker/integrations/registry.rb
 - lib/llm_cost_tracker/llm_api_call.rb
 - lib/llm_cost_tracker/logging.rb
 - lib/llm_cost_tracker/middleware/faraday.rb
@@ -312,31 +332,27 @@ files:
 - lib/llm_cost_tracker/parsers/sse.rb
 - lib/llm_cost_tracker/period_grouping.rb
 - lib/llm_cost_tracker/period_total.rb
+- lib/llm_cost_tracker/price_freshness.rb
 - lib/llm_cost_tracker/price_registry.rb
 - lib/llm_cost_tracker/price_sync.rb
 - lib/llm_cost_tracker/price_sync/fetcher.rb
-- lib/llm_cost_tracker/price_sync/merger.rb
-- lib/llm_cost_tracker/price_sync/model_catalog.rb
-- lib/llm_cost_tracker/price_sync/raw_price.rb
-- lib/llm_cost_tracker/price_sync/refresh_plan_builder.rb
+- lib/llm_cost_tracker/price_sync/registry_diff.rb
 - lib/llm_cost_tracker/price_sync/registry_loader.rb
 - lib/llm_cost_tracker/price_sync/registry_writer.rb
-- lib/llm_cost_tracker/price_sync/source.rb
-- lib/llm_cost_tracker/price_sync/source_result.rb
-- lib/llm_cost_tracker/price_sync/sources/litellm.rb
-- lib/llm_cost_tracker/price_sync/sources/open_router.rb
-- lib/llm_cost_tracker/price_sync/validator.rb
 - lib/llm_cost_tracker/prices.json
 - lib/llm_cost_tracker/pricing.rb
 - lib/llm_cost_tracker/railtie.rb
 - lib/llm_cost_tracker/report.rb
 - lib/llm_cost_tracker/report_data.rb
 - lib/llm_cost_tracker/report_formatter.rb
+- lib/llm_cost_tracker/request_url.rb
 - lib/llm_cost_tracker/retention.rb
 - lib/llm_cost_tracker/storage/active_record_rollups.rb
 - lib/llm_cost_tracker/storage/active_record_store.rb
+- lib/llm_cost_tracker/storage/dispatcher.rb
 - lib/llm_cost_tracker/stream_collector.rb
 - lib/llm_cost_tracker/tag_accessors.rb
+- lib/llm_cost_tracker/tag_context.rb
 - lib/llm_cost_tracker/tag_key.rb
 - lib/llm_cost_tracker/tag_query.rb
 - lib/llm_cost_tracker/tags_column.rb

data/lib/llm_cost_tracker/generators/llm_cost_tracker/templates/llm_cost_tracker_prices.yml.erb

@@ -1,51 +0,0 @@
-# Local LlmCostTracker price overrides.
-#
-# Add only the models you want to override or add. Built-in prices still come
-# from the gem's prices.json, and Ruby pricing_overrides still take precedence.
-#
-# Units: USD per 1M tokens.
-#
-# Supported price keys:
-# - input
-# - output
-# - cache_read_input
-# - cache_write_input
-# - mode_input / mode_output / mode_cache_read_input / mode_cache_write_input
-#
-# Optional metadata keys, ignored by cost calculation:
-# - _source
-# - _source_version
-# - _fetched_at
-# - _updated
-# - _notes
-# - _validator_override
-#
-# Example: custom fine-tune
-# models:
-#   "ft:gpt-4o-mini:my-org":
-#     input: 0.30
-#     cache_read_input: 0.15
-#     output: 1.20
-#     _notes: "Internal fine-tune rate"
-#
-# Example: alternate pricing mode
-# models:
-#   "batchable-model":
-#     input: 1.00
-#     output: 2.00
-#     batch_input: 0.50
-#     batch_output: 1.00
-#
-# Example: negotiated provider discount
-# models:
-#   "gpt-4o":
-#     input: 2.00
-#     output: 8.00
-#     _source: "manual"
-#     _updated: "2026-04-18"
-#
-# Use _source: "manual" for custom or orphaned entries you never want sync to touch.
-# Use _validator_override: ["skip_relative_change"] if a negotiated price would
-# otherwise trip the >3x sync warning.
-
-models:

data/lib/llm_cost_tracker/price_sync/merger.rb

@@ -1,72 +0,0 @@
-# frozen_string_literal: true
-
-module LlmCostTracker
-  module PriceSync
-    class Merger
-      Discrepancy = Data.define(:model, :field, :values)
-
-      PRIORITY_ORDER = %i[litellm openrouter].freeze
-      SUPPLEMENTAL_FIELDS = %i[cache_read_input cache_write_input].freeze
-
-      def merge(results_by_source)
-        prices = collect_prices(results_by_source)
-        discrepancies = []
-
-        merged = prices.group_by(&:model).sort.to_h.transform_values do |candidates|
-          sorted = sort_candidates(candidates)
-          discrepancies.concat(detect_discrepancies(sorted))
-          fill_missing_fields(sorted.first, sorted.drop(1))
-        end
-
-        [merged, discrepancies]
-      end
-
-      private
-
-      def collect_prices(results_by_source)
-        results_by_source.flat_map do |source_name, result|
-          result.prices.map do |price|
-            price.with(source: source_name)
-          end
-        end
-      end
-
-      def sort_candidates(candidates)
-        candidates.sort_by do |price|
-          PRIORITY_ORDER.index(price.source.to_sym) || PRIORITY_ORDER.length
-        end
-      end
-
-      def fill_missing_fields(primary, fallbacks)
-        SUPPLEMENTAL_FIELDS.reduce(primary) do |current, field|
-          next current if current.public_send(field)
-
-          fallback = fallbacks.find { |candidate| candidate.public_send(field) }
-          fallback ? current.with(field => fallback.public_send(field)) : current
-        end
-      end
-
-      def detect_discrepancies(candidates)
-        return [] if candidates.length < 2
-
-        RawPrice::PRICE_FIELDS.filter_map do |field|
-          values = candidates.each_with_object({}) do |price, collected|
-            value = price.public_send(field)
-            collected[price.source] = value unless value.nil?
-          end
-          next if values.size < 2
-          next unless discrepant?(values.values)
-
-          Discrepancy.new(model: candidates.first.model, field: field, values: values)
-        end
-      end
-
-      def discrepant?(values)
-        min, max = values.minmax
-        return max != min if min.to_f.zero?
-
-        ((max - min).abs / min.to_f) >= 0.05
-      end
-    end
-  end
-end