lex-llm-openai 0.3.7 → 0.3.9

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: f49ee94d59da872cc2bb8095bbe619329caecf38f178f1b4c183c626f1f8168c
-  data.tar.gz: 674527b56d8bb6c72a0052bdad2254356e635926b37a1810bea56563f94418be
+  metadata.gz: 808223596350fde35cbbf0145a411842ce1c0a3009d431edd354ce500732666c
+  data.tar.gz: 9ed2580c07f4d4b1c35f9b90e8108b4ddbd9f74fa032998447b3639dda67e4db
 SHA512:
-  metadata.gz: 7dd8317307365087f1279f25f3c263d5dc57dc7c73487f9073d6c77d8a542f335aaf3ed2a4b176f9aee9a1ec59bd5f6269a2cf27845dff9cb73dd8ddbb898833
-  data.tar.gz: 88c350e3870a694eabde6c04580191bc11dcc0f48171df3d37c31f91b8b0c97800a411d78f65d0856304e7710b5420f83ee66514c230ba3a22f75f9afe0b062b
+  metadata.gz: a07fffb4a8a5652e3b14303102bc2811915efc374b32369f50e1d39a6397d44a8a2b7822aebedd300af37cecb1d447cbf3b517fdf8502138077de149c9e9e9ea
+  data.tar.gz: 51f75db914319c8c086d8ca63a0127b6ed79d21ee92dd1a3c25c45bc9082ba0b1fde55792f8e6040e20c247baad09147634b079d953606e3a8031063422ad7f2

data/.rubocop.yml

@@ -14,6 +14,8 @@ Metrics/BlockLength:
     - spec/**/*
 Metrics/ClassLength:
   Max: 200
+Metrics/ModuleLength:
+  Max: 110
 Metrics/MethodLength:
   Enabled: false
 RSpec/ExampleLength:

data/CHANGELOG.md

@@ -1,5 +1,19 @@
 # Changelog
 
+## 0.3.9 - 2026-05-13
+
+- Change `default_model` from `gpt-4o` to `gpt-5.5` in provider default settings and instance discovery fallback.
+- Inject `default_model` into all discovered provider instances so every instance has an explicit model default.
+- Add `context_window` to all `CAPABILITY_MAP` entries (gpt-4o=128K, gpt-4.1/gpt-5=1M, o3/o4/o1=200K, text-embedding=8K).
+- Override `fetch_model_detail` to return `context_window` from the capability map instead of issuing a live API call.
+- Use `model_detail` in `build_model_infos` to populate `context_length` from the cached capability map entry.
+
+## 0.3.8 - 2026-05-13
+
+- Route OpenAI fleet runner and actor diagnostics through `Legion::Logging::Helper` with debug-level request and enablement context.
+- Report optional actor subscription load failures through `handle_exception` instead of raw warning output.
+- Move routine OpenAI model discovery telemetry to debug-level logging while keeping failure handling structured.
+
 ## 0.3.7 - 2026-05-08
 
 - Accept keyword arguments in `list_models` to match the base provider contract called by `discover_offerings`.

data/lib/legion/extensions/llm/openai/actors/fleet_worker.rb

@@ -1,18 +1,25 @@
 # frozen_string_literal: true
 
-begin
-  require 'legion/extensions/actors/subscription'
-rescue LoadError => e
-  warn(e.message) if $VERBOSE
+require 'legion/extensions/llm/openai'
+require 'legion/extensions/llm/fleet/provider_responder'
+
+unless defined?(Legion::Extensions::Actors::Subscription)
+  begin
+    require 'legion/extensions/actors/subscription'
+  rescue LoadError => e
+    Legion::Extensions::Llm::Openai.handle_exception(
+      e,
+      level: :warn,
+      handled: true,
+      operation: 'openai.fleet_worker.load_subscription'
+    )
+  end
 end
 
 unless defined?(Legion::Extensions::Actors::Subscription)
   raise LoadError, 'LegionIO actor runtime is required for OpenAI fleet worker'
 end
 
-require 'legion/extensions/llm/openai'
-require 'legion/extensions/llm/fleet/provider_responder'
-
 module Legion
   module Extensions
     module Llm
@@ -20,6 +27,8 @@ module Legion
         module Actor
           # Subscription actor for OpenAI fleet request consumption.
           class FleetWorker < Legion::Extensions::Actors::Subscription
+            include Legion::Logging::Helper
+
             def runner_class
               'Legion::Extensions::Llm::Openai::Runners::FleetWorker'
             end
@@ -33,7 +42,13 @@ module Legion
             end
 
             def enabled?
-              Legion::Extensions::Llm::Fleet::ProviderResponder.enabled_for?(Openai.discover_instances)
+              instances = Openai.discover_instances
+              enabled = Legion::Extensions::Llm::Fleet::ProviderResponder.enabled_for?(instances)
+              log.debug { "OpenAI fleet worker enablement: enabled=#{enabled}, instance_count=#{instances.size}" }
+              enabled
+            rescue StandardError => e
+              handle_exception(e, level: :warn, handled: true, operation: 'openai.fleet_worker.enabled')
+              false
             end
           end
         end

data/lib/legion/extensions/llm/openai/provider.rb

@@ -19,42 +19,50 @@ module Legion
             'gpt-4o' => {
               capabilities: %i[completion streaming function_calling vision structured_output],
               modalities_input: %w[text image audio],
-              modalities_output: %w[text]
+              modalities_output: %w[text],
+              context_window: 128_000
             },
             'gpt-4.1' => {
               capabilities: %i[completion streaming function_calling vision structured_output],
               modalities_input: %w[text image],
-              modalities_output: %w[text]
+              modalities_output: %w[text],
+              context_window: 1_047_576
             },
             'gpt-4' => {
               capabilities: %i[completion streaming function_calling vision],
               modalities_input: %w[text image],
-              modalities_output: %w[text]
+              modalities_output: %w[text],
+              context_window: 128_000
             },
             'gpt-5' => {
               capabilities: %i[completion streaming function_calling vision structured_output reasoning],
               modalities_input: %w[text image],
-              modalities_output: %w[text]
+              modalities_output: %w[text],
+              context_window: 1_047_576
             },
             'o4' => {
               capabilities: %i[completion streaming function_calling vision reasoning],
               modalities_input: %w[text image],
-              modalities_output: %w[text]
+              modalities_output: %w[text],
+              context_window: 200_000
             },
             'o3' => {
               capabilities: %i[completion streaming function_calling vision reasoning],
               modalities_input: %w[text image],
-              modalities_output: %w[text]
+              modalities_output: %w[text],
+              context_window: 200_000
             },
             'o1' => {
               capabilities: %i[completion streaming function_calling vision reasoning],
               modalities_input: %w[text image],
-              modalities_output: %w[text]
+              modalities_output: %w[text],
+              context_window: 200_000
             },
             'text-embedding-' => {
               capabilities: %i[embedding],
               modalities_input: %w[text],
-              modalities_output: %w[embeddings]
+              modalities_output: %w[embeddings],
+              context_window: 8_191
             },
             'omni-moderation' => {
               capabilities: %i[moderation],
@@ -172,7 +180,7 @@ module Legion
           def images_url(with: nil, mask: nil) = super
 
           def retrieve_model(model)
-            log.info("Retrieving model: #{model}")
+            log.debug { "Retrieving OpenAI model: #{model}" }
             connection.get("#{models_url}/#{model}").body
           rescue StandardError => e
             handle_exception(e, level: :error, handled: true,
@@ -181,10 +189,10 @@ module Legion
           end
 
           def list_models(**)
-            log.info('Listing OpenAI models')
+            log.debug('Listing OpenAI models')
             raw = connection.get(models_url)
             models = build_model_infos(raw.body)
-            log.info("Discovered #{models.size} OpenAI models")
+            log.debug { "Discovered #{models.size} OpenAI models; publishing registry availability" }
             self.class.registry_publisher.publish_models_async(models, readiness: readiness(live: false))
             models
           rescue StandardError => e
@@ -199,12 +207,15 @@ module Legion
             body.fetch('data', []).map do |raw_model|
               id = raw_model.fetch('id')
               cap_entry = capability_entry_for(id)
+              detail = model_detail(id)
+              ctx = detail&.dig(:context_window) || cap_entry[:context_window]
 
               Legion::Extensions::Llm::Model::Info.new(
                 id: id,
                 name: id,
                 provider: :openai,
                 capabilities: cap_entry[:capabilities],
+                context_length: ctx,
                 modalities_input: cap_entry[:modalities_input],
                 modalities_output: cap_entry[:modalities_output],
                 metadata: {
@@ -220,7 +231,6 @@ module Legion
               return entry if model_id.start_with?(prefix)
             end
 
-            # Fallback for unknown models: assume chat-capable
             {
               capabilities: %i[completion streaming],
               modalities_input: %w[text],
@@ -228,6 +238,12 @@ module Legion
             }
           end
 
+          def fetch_model_detail(model_name)
+            entry = capability_entry_for(model_name)
+            ctx = entry[:context_window]
+            ctx ? { context_window: ctx } : nil
+          end
+
           def model_created_at(value)
             value.is_a?(Numeric) ? Time.at(value).utc : value
           end

data/lib/legion/extensions/llm/openai/runners/fleet_worker.rb

@@ -10,9 +10,16 @@ module Legion
         module Runners
           # Runner entrypoint for OpenAI fleet request execution.
           module FleetWorker
+            extend Legion::Logging::Helper
+
             module_function
 
             def handle_fleet_request(payload, delivery: nil, properties: nil)
+              log.debug do
+                "Handling OpenAI fleet request: request_id=#{payload_value(payload, :request_id) || 'unknown'}, " \
+                  "provider_instance=#{payload_value(payload, :provider_instance) || 'default'}"
+              end
+
               Legion::Extensions::Llm::Fleet::ProviderResponder.call(
                 payload: payload,
                 provider_family: Openai::PROVIDER_FAMILY,
@@ -21,6 +28,15 @@ module Legion
                 delivery: delivery,
                 properties: properties
               )
+            rescue StandardError => e
+              handle_exception(e, level: :error, handled: true, operation: 'openai.fleet_worker.handle_request')
+              raise
+            end
+
+            def payload_value(payload, key)
+              return unless payload.respond_to?(:[])
+
+              payload[key] || payload[key.to_s]
             end
           end
         end

data/lib/legion/extensions/llm/openai/version.rb

@@ -4,7 +4,7 @@ module Legion
   module Extensions
     module Llm
       module Openai
-        VERSION = '0.3.7'
+        VERSION = '0.3.9'
       end
     end
   end

data/lib/legion/extensions/llm/openai.rb

@@ -20,7 +20,7 @@ module Legion
             family: PROVIDER_FAMILY,
             instance: {
               endpoint: 'https://api.openai.com',
-              default_model: 'gpt-4o',
+              default_model: 'gpt-5.5',
               tier: :frontier,
               transport: :http,
               credentials: {
@@ -53,6 +53,7 @@ module Legion
         end
 
         def self.discover_instances # rubocop:disable Metrics/AbcSize,Metrics/CyclomaticComplexity,Metrics/PerceivedComplexity
+          log.debug('Discovering OpenAI provider instances')
           candidates = {}
 
           # 1. OPENAI_API_KEY environment variable
@@ -100,8 +101,15 @@ module Legion
             candidates[name.to_sym] = normalized.merge(tier: :frontier)
           end
 
-          # 8. Dedup
-          CredentialSources.dedup_credentials(candidates).transform_values { |config| sanitize_instance_config(config) }
+          # 8. Dedup + inject default_model
+          discovered = CredentialSources.dedup_credentials(candidates).transform_values do |config|
+            sanitized = sanitize_instance_config(config)
+            sanitized[:default_model] ||= 'gpt-5.5'
+            sanitized
+          end
+          instance_names = discovered.keys.sort_by(&:to_s).join(', ')
+          log.debug { "Discovered #{discovered.size} OpenAI provider instance candidate(s): #{instance_names}" }
+          discovered
         end
 
         def self.settings_instances(config)

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: lex-llm-openai
 version: !ruby/object:Gem::Version
-  version: 0.3.7
+  version: 0.3.9
 platform: ruby
 authors:
 - LegionIO