lex-llm-vllm 0.2.0 → 0.2.6

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 261a83f8a4243c795e10e759e5f7f0681cd02d91b91041ff6f11f50d498d1f2a
-  data.tar.gz: c136080255383ca0d3417937d3e4988701075fef4df4c33de3e85fbd9f6ae04b
+  metadata.gz: 14dc7ee5334135f8eece7622a27cb57bad3c9c885cb6203de56006b5d2a1b353
+  data.tar.gz: 59ea84f7c50a9407da2af50c51f77ad26894d45b7f06d9c0223198a73a7e22a8
 SHA512:
-  metadata.gz: a521587328074e46f4403783b85d001f6a9a4cab77e31556ea565432dea12535ed9d0c656b8f8212b20d37f196d8c08f86be25955ff25813ca908e03b5fa8e60
-  data.tar.gz: 3b74fa8c6ecfd4c71fb027a5eb13d3d41b7acf1c42cebcbfeea0e4143ec5ef9ad2693020c284eb3377a8f77fe6cd492c85ed9b9af8f515ad32918f152df224a8
+  metadata.gz: 05c8fa912c908ec88943277dfd2e2b82f84bbd4ff6880d4e1f288f1397baf2cc1f72c2d20e894684be14be9f5f213e381ea5f524bfed3162e116de0271009f93
+  data.tar.gz: 14fffc18eff78d0c8751fbb96c9db6cbdd2773af66c46359b5934fad09d9a8fa37183a4f123f7cb75008e44cf821a21d97756d47633e44a50ab935f3f5743f0d

data/.github/workflows/ci.yml

@@ -8,8 +8,20 @@ jobs:
   ci:
     uses: LegionIO/.github/.github/workflows/ci.yml@main
 
+  excluded-files:
+    uses: LegionIO/.github/.github/workflows/excluded-files.yml@main
+
+  security:
+    uses: LegionIO/.github/.github/workflows/security-scan.yml@main
+
+  version-changelog:
+    uses: LegionIO/.github/.github/workflows/version-changelog.yml@main
+
+  dependency-review:
+    uses: LegionIO/.github/.github/workflows/dependency-review.yml@main
+
   release:
-    needs: ci
+    needs: [ci, excluded-files, security]
     if: github.event_name == 'push' && github.ref == 'refs/heads/main'
     uses: LegionIO/.github/.github/workflows/release.yml@main
     secrets:

data/CHANGELOG.md

@@ -1,5 +1,42 @@
 # Changelog
 
+## 0.2.6 - 2026-05-06
+
+- Load provider-owned fleet actors through the LegionIO subscription base and the canonical vLLM provider root.
+- Keep fleet runners anchored on the provider root namespace so provider constants and instance discovery are always loaded.
+- Normalize configured `endpoint` and `api_base` aliases to `vllm_api_base`.
+- Preserve configured transport and tier metadata when vLLM builds routing offerings.
+- Gate release publishing on the shared security workflow.
+
+## 0.2.5 - 2026-05-06
+
+- Mark handled vLLM offering-discovery failures as handled when logging through `Legion::Logging::Helper`.
+- Refresh README dependency, defaults, and local verification guidance for the `lex-llm >= 0.4.3` fleet responder contract.
+
+## 0.2.4 - 2026-05-06
+
+- Use the shared `lex-llm` fleet provider responder helper for provider-owned fleet workers.
+- Remove the runtime `legion-llm` dependency and require `lex-llm >= 0.4.3` for responder-side fleet execution.
+
+## 0.2.3 - 2026-05-06
+
+- Remove require-time provider self-registration; `legion-llm` now owns adapter creation and registry writes from loaded provider discovery metadata.
+- Bump dependency floors to `lex-llm >= 0.4.1` and `legion-llm >= 0.9.1`.
+
+## 0.2.2 - 2026-05-06
+
+- Enforce the shared keyword-only `lex-llm` provider contract and accept `health(live:)`.
+- Move vLLM defaults back to `Legion::Extensions::Llm.provider_settings` with instance-level fleet responder settings.
+- Read vLLM thinking defaults from the nested provider instance settings shape.
+- Serve non-live vLLM offering reads from cached live model discovery instead of probing the configured endpoint.
+- Add provider-owned fleet responder actor and runner backed by `legion-llm` fleet policy execution.
+- Bump the transport dependency floor to `legion-transport >= 1.4.14`.
+
+## 0.2.1 - 2026-05-03
+
+- Normalize configured `base_url` instance settings to `vllm_api_base` so LegionIO local settings are honored during provider registration.
+- Strip a trailing `/v1` from configured vLLM API roots because OpenAI-compatible endpoints append their own `/v1/...` paths.
+
 ## 0.2.0 - 2026-05-01
 
 - Add auto-discovery via CredentialSources and AutoRegistration from lex-llm 0.3.0

data/Gemfile

@@ -4,6 +4,8 @@ source 'https://rubygems.org'
 
 group :test do
   llm_base_path = ENV.fetch('LEX_LLM_PATH', File.expand_path('../lex-llm', __dir__))
+  transport_path = ENV.fetch('LEGION_TRANSPORT_PATH', File.expand_path('../../legion-transport', __dir__))
+  gem 'legion-transport', path: transport_path if File.directory?(transport_path)
   gem 'lex-llm', path: llm_base_path if File.directory?(llm_base_path)
 end
 

data/README.md

@@ -2,7 +2,7 @@
 
 LegionIO LLM provider extension for [vLLM](https://docs.vllm.ai/).
 
-This gem lives under `Legion::Extensions::Llm::Vllm` and depends on `lex-llm` for shared provider-neutral routing, fleet, and schema primitives.
+This gem lives under `Legion::Extensions::Llm::Vllm` and depends on `lex-llm >= 0.4.3` for shared provider-neutral routing, response normalization, fleet envelopes, responder-side fleet execution, and schema primitives.
 
 Load it with `require 'legion/extensions/llm/vllm'`.
 
@@ -19,7 +19,7 @@ Load it with `require 'legion/extensions/llm/vllm'`.
 - vLLM management helpers: `/health`, `/version`, `/reset_prefix_cache`, `/reset_mm_cache`, `/sleep`, `/wake_up`
 - Normalized OpenAI-compatible capability and modality metadata for discovered models
 - Shared fleet/default settings via `Legion::Extensions::Llm.provider_settings`
-- Full `Legion::Logging::Helper` integration with structured `handle_exception` across all classes
+- Structured `Legion::Logging::Helper` handling for provider discovery and fallback paths
 
 ## Defaults
 
@@ -30,10 +30,20 @@ Legion::Extensions::Llm::Vllm.default_settings
 #   instances: {
 #     default: {
 #       endpoint: "http://localhost:8000",
-#       tier: :private,
+#       tier: :direct,
 #       transport: :http,
-#       usage: { inference: true, embedding: true },
-#       limits: { concurrency: 8 }
+#       credentials: { api_key: nil },
+#       enable_thinking: true,
+#       usage: { inference: true, embedding: true, image: true },
+#       limits: { concurrency: 1 },
+#       fleet: {
+#         enabled: false,
+#         respond_to_requests: false,
+#         capabilities: [:chat, :stream_chat, :embed],
+#         lanes: [],
+#         concurrency: 1,
+#         queue_suffix: nil
+#       }
 #     }
 #   }
 # }
@@ -50,6 +60,25 @@ Legion::Extensions::Llm.configure do |config|
 end
 ```
 
+## Fleet Responder
+
+Provider instances can opt in to consuming Legion LLM fleet requests. The provider-owned fleet actor only starts when at least one configured instance enables `respond_to_requests`, and request execution delegates to `Legion::Extensions::Llm::Fleet::ProviderResponder`.
+
+```yaml
+extensions:
+  llm:
+    vllm:
+      instances:
+        local:
+          fleet:
+            enabled: true
+            respond_to_requests: true
+            capabilities:
+              - chat
+              - stream_chat
+              - embed
+```
+
 ### Thinking Mode
 
 Enable vLLM thinking mode globally via settings:
@@ -87,8 +116,8 @@ Publishing is async (background threads) and never blocks the caller. All failur
 
 ```bash
 bundle install
-bundle exec rspec
-bundle exec rubocop
+bundle exec rspec --format json --out tmp/rspec_results.json --format progress --out tmp/rspec_progress.txt
+bundle exec rubocop -A
 ```
 
 ## License

data/lex-llm-vllm.gemspec

@@ -26,5 +26,6 @@ Gem::Specification.new do |spec|
   spec.add_dependency 'legion-json', '>= 1.2.1'
   spec.add_dependency 'legion-logging', '>= 1.3.2'
   spec.add_dependency 'legion-settings', '>= 1.3.14'
-  spec.add_dependency 'lex-llm', '>= 0.3.0'
+  spec.add_dependency 'legion-transport', '>= 1.4.14'
+  spec.add_dependency 'lex-llm', '>= 0.4.3'
 end

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

@@ -0,0 +1,43 @@
+# frozen_string_literal: true
+
+begin
+  require 'legion/extensions/actors/subscription'
+rescue LoadError => e
+  warn(e.message) if $VERBOSE
+end
+
+unless defined?(Legion::Extensions::Actors::Subscription)
+  raise LoadError, 'LegionIO actor runtime is required for vLLM fleet worker'
+end
+
+require 'legion/extensions/llm/vllm'
+require 'legion/extensions/llm/fleet/provider_responder'
+
+module Legion
+  module Extensions
+    module Llm
+      module Vllm
+        module Actor
+          # Subscription actor for vLLM fleet request consumption.
+          class FleetWorker < Legion::Extensions::Actors::Subscription
+            def runner_class
+              'Legion::Extensions::Llm::Vllm::Runners::FleetWorker'
+            end
+
+            def runner_function
+              'handle_fleet_request'
+            end
+
+            def use_runner?
+              false
+            end
+
+            def enabled?
+              Legion::Extensions::Llm::Fleet::ProviderResponder.enabled_for?(Vllm.discover_instances)
+            end
+          end
+        end
+      end
+    end
+  end
+end

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

@@ -68,8 +68,8 @@ module Legion
           def sleep_url = '/sleep'
           def wake_up_url = '/wake_up'
 
-          def health
-            log.info { "checking health at #{api_base}#{health_url}" }
+          def health(live: false)
+            log.info { "checking health live=#{live} at #{api_base}#{health_url}" }
             connection.get(health_url).body
           end
 
@@ -88,6 +88,18 @@ module Legion
             end
           end
 
+          def discover_offerings(live: false, **)
+            models = if live
+                       @cached_models = list_models
+                     else
+                       Array(@cached_models)
+                     end
+            models.map { |model_info| offering_from_model(model_info) }
+          rescue StandardError => e
+            handle_exception(e, level: :warn, handled: true, operation: 'vllm.discover_offerings')
+            []
+          end
+
           def version
             log.info { "fetching version from #{api_base}#{version_url}" }
             connection.get(version_url).body
@@ -112,6 +124,28 @@ module Legion
 
           private
 
+          def offering_from_model(model_info)
+            Legion::Extensions::Llm::Routing::ModelOffering.new(
+              provider_family: :vllm,
+              instance_id: config.respond_to?(:instance_id) ? config.instance_id : :default,
+              transport: offering_transport,
+              tier: offering_tier,
+              model: model_info.id,
+              usage_type: model_info.embedding? ? :embedding : :inference,
+              capabilities: model_info.capabilities.map(&:to_s),
+              limits: { context_window: model_info.context_length }.compact,
+              metadata: { context_length: model_info.context_length }
+            )
+          end
+
+          def offering_transport
+            config.respond_to?(:transport) ? config.transport : :http
+          end
+
+          def offering_tier
+            config.respond_to?(:tier) ? config.tier : :direct
+          end
+
           def render_payload(messages, tools:, temperature:, model:, stream:, schema:, thinking:, tool_prefs:) # rubocop:disable Metrics/ParameterLists
             payload = super
             payload.delete(:reasoning_effort)
@@ -131,7 +165,12 @@ module Legion
             return false unless defined?(Legion::Settings)
 
             vllm = Legion::Settings.dig(:llm, :providers, :vllm)
-            vllm.is_a?(Hash) && (vllm[:enable_thinking] == true || vllm['enable_thinking'] == true)
+            return false unless vllm.is_a?(Hash)
+
+            vllm[:enable_thinking] == true ||
+              vllm['enable_thinking'] == true ||
+              vllm.dig(:instances, :default, :enable_thinking) == true ||
+              vllm.dig('instances', 'default', 'enable_thinking') == true
           rescue StandardError => e
             handle_exception(e, level: :debug, handled: true, operation: 'vllm.thinking_setting')
             false

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

@@ -0,0 +1,30 @@
+# frozen_string_literal: true
+
+require 'legion/extensions/llm/fleet/provider_responder'
+require 'legion/extensions/llm/vllm'
+
+module Legion
+  module Extensions
+    module Llm
+      module Vllm
+        module Runners
+          # Runner entrypoint for vLLM fleet request execution.
+          module FleetWorker
+            module_function
+
+            def handle_fleet_request(payload, delivery: nil, properties: nil)
+              Legion::Extensions::Llm::Fleet::ProviderResponder.call(
+                payload: payload,
+                provider_family: Vllm::PROVIDER_FAMILY,
+                provider_class: Vllm::Provider,
+                provider_instances: -> { Vllm.discover_instances },
+                delivery: delivery,
+                properties: properties
+              )
+            end
+          end
+        end
+      end
+    end
+  end
+end

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

@@ -4,7 +4,7 @@ module Legion
   module Extensions
     module Llm
       module Vllm
-        VERSION = '0.2.0'
+        VERSION = '0.2.6'
       end
     end
   end

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

@@ -16,17 +16,26 @@ module Legion
         PROVIDER_FAMILY = :vllm
 
         def self.default_settings
-          {
-            enabled: false,
-            base_url: 'localhost:8000/v1',
-            default_model: nil,
-            enable_thinking: true,
-            model_whitelist: [],
-            model_blacklist: [],
-            model_cache_ttl: 300,
-            tls: { enabled: false, verify: :peer },
-            instances: {}
-          }
+          ::Legion::Extensions::Llm.provider_settings(
+            family: PROVIDER_FAMILY,
+            instance: {
+              endpoint: 'http://localhost:8000',
+              tier: :direct,
+              transport: :http,
+              credentials: { api_key: nil },
+              enable_thinking: true,
+              usage: { inference: true, embedding: true, image: true },
+              limits: { concurrency: 1 },
+              fleet: {
+                enabled: false,
+                respond_to_requests: false,
+                capabilities: %i[chat stream_chat embed],
+                lanes: [],
+                concurrency: 1,
+                queue_suffix: nil
+              }
+            }
+          )
         end
 
         def self.provider_class
@@ -51,19 +60,26 @@ module Legion
           configured = CredentialSources.setting(:extensions, :llm, :vllm, :instances)
           if configured.is_a?(Hash)
             configured.each do |name, config|
-              instances[name.to_sym] = config.merge(tier: :direct)
+              instances[name.to_sym] = normalize_instance_config(config).merge(tier: :direct)
             end
           end
 
           instances
         end
 
-        if Legion::Extensions::Llm::Configuration.respond_to?(:register_provider_options)
-          Legion::Extensions::Llm::Configuration.register_provider_options(Provider.configuration_options)
+        def self.normalize_instance_config(config)
+          normalized = config.to_h.transform_keys(&:to_sym)
+          normalized[:vllm_api_base] ||= normalized.delete(:base_url)
+          normalized[:vllm_api_base] ||= normalized.delete(:api_base)
+          normalized[:vllm_api_base] ||= normalized.delete(:endpoint)
+          normalized[:vllm_api_base] = normalize_api_base(normalized[:vllm_api_base]) if normalized[:vllm_api_base]
+          normalized
+        end
+
+        def self.normalize_api_base(url)
+          url.to_s.sub(%r{/v1/?\z}, '')
         end
       end
     end
   end
 end
-
-Legion::Extensions::Llm::Vllm.register_discovered_instances

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: lex-llm-vllm
 version: !ruby/object:Gem::Version
-  version: 0.2.0
+  version: 0.2.6
 platform: ruby
 authors:
 - LegionIO
@@ -51,20 +51,34 @@ dependencies:
     - - ">="
       - !ruby/object:Gem::Version
         version: 1.3.14
+- !ruby/object:Gem::Dependency
+  name: legion-transport
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: 1.4.14
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: 1.4.14
 - !ruby/object:Gem::Dependency
   name: lex-llm
   requirement: !ruby/object:Gem::Requirement
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: 0.3.0
+        version: 0.4.3
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: 0.3.0
+        version: 0.4.3
 description: vLLM provider integration for the LegionIO LLM routing framework.
 email:
 - matthewdiverson@gmail.com
@@ -83,7 +97,9 @@ files:
 - README.md
 - lex-llm-vllm.gemspec
 - lib/legion/extensions/llm/vllm.rb
+- lib/legion/extensions/llm/vllm/actors/fleet_worker.rb
 - lib/legion/extensions/llm/vllm/provider.rb
+- lib/legion/extensions/llm/vllm/runners/fleet_worker.rb
 - lib/legion/extensions/llm/vllm/version.rb
 homepage: https://github.com/LegionIO/lex-llm-vllm
 licenses: