lex-llm-vllm 0.1.3 → 0.1.6

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 CHANGED
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  SHA256:
3
- metadata.gz: dca0ffb2e1335cffd3b857e687ff1d0bb51846af7f9bcaaa108a61a227058286
4
- data.tar.gz: 71811d068ce3df025d7e83225ff048c7e75f293004565531287678601468aac1
3
+ metadata.gz: 3d9ce01c4d82eba349cdbe37aca5fa4b577dfc7bcf8d2b96c54353760f2162b0
4
+ data.tar.gz: dfe94103d8c4b71a7af0b5108e4bb021080406f065ad4cf16ae99ea67f696c16
5
5
  SHA512:
6
- metadata.gz: 82427659e263f7427c0371e4a3b53af87bc7f93242038e61ea249e3a1fb5b4ef1d5c89b0689c82102b294bfd76bd7453f4e8b795216fe2452c5756af6aad56cd
7
- data.tar.gz: 88688cb6c5a56bdbc59e4ae5189816c0979d38cf37f1fc70ad2f3628d3939e32bf313efffa4eae81ac917bc18c6c2a597ad051906410c2c07144ee8dfb912858
6
+ metadata.gz: e0dff8aa6b171711dde2fc016f0a899e56b3ee54ed83868a19c8cd9420f7ff9b8e16cb2164df41cda3ae2c6d3e0c0400591a04dc6308abeca72ed31ec0627241
7
+ data.tar.gz: 283a82ff2cad5658fa7250e7c3a2a0492dd0f7d2de471352a89449b8874957f2f8062265b4497f46a43e83e0308458b64d01915d7d170e240ab0f5ffddaf50e9
data/CHANGELOG.md CHANGED
@@ -1,5 +1,18 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.1.6 - 2026-04-28
4
+
5
+ - Publish best-effort `llm.registry` readiness and discovered-model availability events when transport is loaded.
6
+
7
+ ## 0.1.5 - 2026-04-28
8
+
9
+ - Require current shared Legion JSON, logging, settings, and LLM extension gems.
10
+
11
+ ## 0.1.4 - 2026-04-28
12
+
13
+ - Require `lex-llm >= 0.1.4` so OpenAI-compatible model discovery exposes normalized capabilities and modalities.
14
+ - Add explicit discovered-model capability mapping for vLLM routing metadata.
15
+
3
16
  ## 0.1.3 - 2026-04-28
4
17
 
5
18
  - Remove the leftover compatibility entrypoint outside the Legion namespace.
data/README.md CHANGED
@@ -15,6 +15,7 @@ Load it with `require 'legion/extensions/llm/vllm'`.
15
15
  - model discovery through `GET /v1/models`
16
16
  - embeddings through `POST /v1/embeddings`
17
17
  - vLLM management helpers for `/health`, `/version`, `/reset_prefix_cache`, `/reset_mm_cache`, `/sleep`, and `/wake_up`
18
+ - normalized OpenAI-compatible capability and modality metadata for discovered models
18
19
  - shared fleet/default settings via `Legion::Extensions::Llm.provider_settings`
19
20
 
20
21
  ## Defaults
data/lex-llm-vllm.gemspec CHANGED
@@ -23,5 +23,8 @@ Gem::Specification.new do |spec|
23
23
  spec.files = `git ls-files -z`.split("\x0").reject { |file| file.match(%r{^(spec|test|features|tmp|coverage)/}) }
24
24
  spec.require_paths = ['lib']
25
25
 
26
- spec.add_dependency 'lex-llm', '>= 0.1.3'
26
+ spec.add_dependency 'legion-json', '>= 1.2.1'
27
+ spec.add_dependency 'legion-logging', '>= 1.3.2'
28
+ spec.add_dependency 'legion-settings', '>= 1.3.14'
29
+ spec.add_dependency 'lex-llm', '>= 0.1.5'
27
30
  end
@@ -12,11 +12,17 @@ module Legion
12
12
  include Legion::Extensions::Llm::Provider::OpenAICompatible
13
13
 
14
14
  class << self
15
+ attr_writer :registry_publisher
16
+
15
17
  def slug = 'vllm'
16
18
  def local? = true
17
19
  def configuration_options = %i[vllm_api_base vllm_api_key]
18
20
  def configuration_requirements = []
19
21
  def capabilities = Capabilities
22
+
23
+ def registry_publisher
24
+ @registry_publisher ||= RegistryPublisher.new
25
+ end
20
26
  end
21
27
 
22
28
  # Capability predicates for vLLM OpenAI-compatible model offerings.
@@ -28,6 +34,15 @@ module Legion
28
34
  def vision?(_model) = true
29
35
  def functions?(_model) = true
30
36
  def embeddings?(_model) = true
37
+
38
+ def critical_capabilities_for(model)
39
+ [
40
+ ('streaming' if streaming?(model)),
41
+ ('function_calling' if functions?(model)),
42
+ ('vision' if vision?(model)),
43
+ ('embeddings' if embeddings?(model))
44
+ ].compact
45
+ end
31
46
  end
32
47
 
33
48
  def api_base
@@ -52,6 +67,18 @@ module Legion
52
67
  connection.get(health_url).body
53
68
  end
54
69
 
70
+ def readiness(live: false)
71
+ super.tap do |metadata|
72
+ self.class.registry_publisher.publish_readiness_async(metadata) if live
73
+ end
74
+ end
75
+
76
+ def list_models
77
+ super.tap do |models|
78
+ self.class.registry_publisher.publish_models_async(models, readiness: readiness(live: false))
79
+ end
80
+ end
81
+
55
82
  def version
56
83
  connection.get(version_url).body
57
84
  end
@@ -0,0 +1,122 @@
1
+ # frozen_string_literal: true
2
+
3
+ module Legion
4
+ module Extensions
5
+ module Llm
6
+ module Vllm
7
+ # Builds sanitized lex-llm registry envelopes for vLLM provider state.
8
+ class RegistryEventBuilder
9
+ def readiness(readiness)
10
+ registry_event_class.public_send(
11
+ readiness[:ready] ? :available : :unavailable,
12
+ provider_offering(readiness),
13
+ runtime: runtime_metadata,
14
+ health: readiness_health(readiness),
15
+ metadata: readiness_metadata(readiness)
16
+ )
17
+ end
18
+
19
+ def model_available(model, readiness:)
20
+ registry_event_class.available(
21
+ model_offering(model),
22
+ runtime: runtime_metadata,
23
+ health: model_health(readiness),
24
+ metadata: model_metadata(model)
25
+ )
26
+ end
27
+
28
+ private
29
+
30
+ def provider_offering(readiness)
31
+ {
32
+ provider_family: :vllm,
33
+ provider_instance: provider_instance,
34
+ transport: :http,
35
+ model: 'provider-readiness',
36
+ usage_type: :inference,
37
+ capabilities: [],
38
+ health: readiness_health(readiness),
39
+ metadata: { lex: :llm_vllm, provider_readiness: true }
40
+ }
41
+ end
42
+
43
+ def model_offering(model)
44
+ {
45
+ provider_family: :vllm,
46
+ provider_instance: provider_instance,
47
+ transport: :http,
48
+ model: model.id,
49
+ usage_type: usage_type_for(model),
50
+ capabilities: Array(model.capabilities).map(&:to_sym),
51
+ limits: model_limits(model),
52
+ metadata: { lex: :llm_vllm, model_name: model.name }.compact
53
+ }
54
+ end
55
+
56
+ def readiness_health(readiness)
57
+ health = {
58
+ ready: readiness[:ready] == true,
59
+ status: readiness[:ready] ? :available : :unavailable,
60
+ checked: readiness.dig(:health, :checked) != false
61
+ }
62
+ add_readiness_error(health, readiness[:health])
63
+ end
64
+
65
+ def add_readiness_error(health, source)
66
+ error = source.is_a?(Hash) ? source : {}
67
+ error_class = error[:error] || error['error']
68
+ error_message = error[:message] || error['message']
69
+ health[:error_class] = error_class if error_class
70
+ health[:error] = error_message if error_message
71
+ health
72
+ end
73
+
74
+ def model_health(readiness)
75
+ ready = readiness.fetch(:ready, true) == true
76
+ { ready:, status: ready ? :available : :degraded }
77
+ end
78
+
79
+ def readiness_metadata(readiness)
80
+ {
81
+ extension: :lex_llm_vllm,
82
+ provider: :vllm,
83
+ configured: readiness[:configured] == true,
84
+ live: readiness[:live] == true
85
+ }
86
+ end
87
+
88
+ def model_metadata(model)
89
+ { extension: :lex_llm_vllm, provider: :vllm, model_type: model.type }
90
+ end
91
+
92
+ def runtime_metadata
93
+ { node: provider_instance }
94
+ end
95
+
96
+ def model_limits(model)
97
+ {
98
+ context_window: model.context_window,
99
+ max_output_tokens: model.max_output_tokens
100
+ }.compact
101
+ end
102
+
103
+ def usage_type_for(model)
104
+ model.type == 'embedding' ? :embedding : :inference
105
+ end
106
+
107
+ def provider_instance
108
+ configured_node = (::Legion::Settings.dig(:node, :canonical_name) if defined?(::Legion::Settings))
109
+ value = configured_node.to_s.strip
110
+ value.empty? ? :vllm : value.to_sym
111
+ rescue StandardError
112
+ :vllm
113
+ end
114
+
115
+ def registry_event_class
116
+ ::Legion::Extensions::Llm::Routing::RegistryEvent
117
+ end
118
+ end
119
+ end
120
+ end
121
+ end
122
+ end
@@ -0,0 +1,100 @@
1
+ # frozen_string_literal: true
2
+
3
+ module Legion
4
+ module Extensions
5
+ module Llm
6
+ module Vllm
7
+ # Best-effort publisher for vLLM provider availability events.
8
+ class RegistryPublisher
9
+ APP_ID = 'lex-llm-vllm'
10
+
11
+ def initialize(builder: RegistryEventBuilder.new)
12
+ @builder = builder
13
+ end
14
+
15
+ def publish_readiness_async(readiness)
16
+ schedule { publish_event(@builder.readiness(readiness)) }
17
+ end
18
+
19
+ def publish_models_async(models, readiness:)
20
+ schedule do
21
+ Array(models).each do |model|
22
+ publish_event(@builder.model_available(model, readiness:))
23
+ end
24
+ end
25
+ end
26
+
27
+ private
28
+
29
+ def schedule(&)
30
+ return false unless publishing_available?
31
+
32
+ Thread.new do
33
+ Thread.current.abort_on_exception = false
34
+ yield
35
+ rescue StandardError => e
36
+ log_publish_failure(e, level: :debug)
37
+ end
38
+ rescue StandardError => e
39
+ log_publish_failure(e, level: :debug)
40
+ false
41
+ end
42
+
43
+ def publish_event(event)
44
+ return false unless publishing_available?
45
+
46
+ message_class.new(event:, app_id: APP_ID).publish(spool: false)
47
+ rescue StandardError => e
48
+ log_publish_failure(e)
49
+ false
50
+ end
51
+
52
+ def publishing_available?
53
+ return false unless registry_event_available?
54
+ return false unless transport_message_available?
55
+ return true unless defined?(::Legion::Transport::Connection)
56
+ return true unless ::Legion::Transport::Connection.respond_to?(:session_open?)
57
+
58
+ ::Legion::Transport::Connection.session_open?
59
+ rescue StandardError
60
+ false
61
+ end
62
+
63
+ def registry_event_available?
64
+ defined?(::Legion::Extensions::Llm::Routing::RegistryEvent)
65
+ end
66
+
67
+ def transport_message_available?
68
+ return true if message_class_defined?
69
+ return false unless defined?(::Legion::Transport::Message) && defined?(::Legion::Transport::Exchange)
70
+
71
+ require 'legion/extensions/llm/vllm/transport/messages/registry_event'
72
+ message_class_defined?
73
+ rescue LoadError
74
+ false
75
+ end
76
+
77
+ def message_class_defined?
78
+ defined?(::Legion::Extensions::Llm::Vllm::Transport::Messages::RegistryEvent)
79
+ end
80
+
81
+ def message_class
82
+ ::Legion::Extensions::Llm::Vllm::Transport::Messages::RegistryEvent
83
+ end
84
+
85
+ def log_publish_failure(error, level: :warn)
86
+ message = "[lex-llm-vllm] llm.registry publish failed: #{error.class}: #{error.message}"
87
+ logger = ::Legion::Extensions::Llm.logger if defined?(::Legion::Extensions::Llm)
88
+ if logger.respond_to?(level)
89
+ logger.public_send(level, message)
90
+ elsif logger.respond_to?(:debug)
91
+ logger.debug(message)
92
+ end
93
+ rescue StandardError
94
+ nil
95
+ end
96
+ end
97
+ end
98
+ end
99
+ end
100
+ end
@@ -0,0 +1,24 @@
1
+ # frozen_string_literal: true
2
+
3
+ module Legion
4
+ module Extensions
5
+ module Llm
6
+ module Vllm
7
+ module Transport
8
+ module Exchanges
9
+ # Topic exchange for vLLM provider availability events.
10
+ class LlmRegistry < ::Legion::Transport::Exchange
11
+ def exchange_name
12
+ 'llm.registry'
13
+ end
14
+
15
+ def default_type
16
+ 'topic'
17
+ end
18
+ end
19
+ end
20
+ end
21
+ end
22
+ end
23
+ end
24
+ end
@@ -0,0 +1,42 @@
1
+ # frozen_string_literal: true
2
+
3
+ require 'legion/extensions/llm/vllm/transport/exchanges/llm_registry'
4
+
5
+ module Legion
6
+ module Extensions
7
+ module Llm
8
+ module Vllm
9
+ module Transport
10
+ module Messages
11
+ # Publishes lex-llm RegistryEvent envelopes to the llm.registry exchange.
12
+ class RegistryEvent < ::Legion::Transport::Message
13
+ def initialize(event:, **options)
14
+ super(**event.to_h.merge(options))
15
+ end
16
+
17
+ def exchange
18
+ Transport::Exchanges::LlmRegistry
19
+ end
20
+
21
+ def routing_key
22
+ @options[:routing_key] || "llm.registry.#{@options.fetch(:event_type)}"
23
+ end
24
+
25
+ def type
26
+ 'llm.registry.event'
27
+ end
28
+
29
+ def app_id
30
+ @options[:app_id] || RegistryPublisher::APP_ID
31
+ end
32
+
33
+ def persistent # rubocop:disable Naming/PredicateMethod
34
+ false
35
+ end
36
+ end
37
+ end
38
+ end
39
+ end
40
+ end
41
+ end
42
+ end
@@ -4,7 +4,7 @@ module Legion
4
4
  module Extensions
5
5
  module Llm
6
6
  module Vllm
7
- VERSION = '0.1.3'
7
+ VERSION = '0.1.6'
8
8
  end
9
9
  end
10
10
  end
@@ -1,7 +1,9 @@
1
1
  # frozen_string_literal: true
2
2
 
3
3
  require 'legion/extensions/llm'
4
+ require 'legion/extensions/llm/vllm/registry_event_builder'
4
5
  require 'legion/extensions/llm/vllm/provider'
6
+ require 'legion/extensions/llm/vllm/registry_publisher'
5
7
  require 'legion/extensions/llm/vllm/version'
6
8
 
7
9
  module Legion
metadata CHANGED
@@ -1,7 +1,7 @@
1
1
  --- !ruby/object:Gem::Specification
2
2
  name: lex-llm-vllm
3
3
  version: !ruby/object:Gem::Version
4
- version: 0.1.3
4
+ version: 0.1.6
5
5
  platform: ruby
6
6
  authors:
7
7
  - LegionIO
@@ -9,20 +9,62 @@ bindir: bin
9
9
  cert_chain: []
10
10
  date: 1980-01-02 00:00:00.000000000 Z
11
11
  dependencies:
12
+ - !ruby/object:Gem::Dependency
13
+ name: legion-json
14
+ requirement: !ruby/object:Gem::Requirement
15
+ requirements:
16
+ - - ">="
17
+ - !ruby/object:Gem::Version
18
+ version: 1.2.1
19
+ type: :runtime
20
+ prerelease: false
21
+ version_requirements: !ruby/object:Gem::Requirement
22
+ requirements:
23
+ - - ">="
24
+ - !ruby/object:Gem::Version
25
+ version: 1.2.1
26
+ - !ruby/object:Gem::Dependency
27
+ name: legion-logging
28
+ requirement: !ruby/object:Gem::Requirement
29
+ requirements:
30
+ - - ">="
31
+ - !ruby/object:Gem::Version
32
+ version: 1.3.2
33
+ type: :runtime
34
+ prerelease: false
35
+ version_requirements: !ruby/object:Gem::Requirement
36
+ requirements:
37
+ - - ">="
38
+ - !ruby/object:Gem::Version
39
+ version: 1.3.2
40
+ - !ruby/object:Gem::Dependency
41
+ name: legion-settings
42
+ requirement: !ruby/object:Gem::Requirement
43
+ requirements:
44
+ - - ">="
45
+ - !ruby/object:Gem::Version
46
+ version: 1.3.14
47
+ type: :runtime
48
+ prerelease: false
49
+ version_requirements: !ruby/object:Gem::Requirement
50
+ requirements:
51
+ - - ">="
52
+ - !ruby/object:Gem::Version
53
+ version: 1.3.14
12
54
  - !ruby/object:Gem::Dependency
13
55
  name: lex-llm
14
56
  requirement: !ruby/object:Gem::Requirement
15
57
  requirements:
16
58
  - - ">="
17
59
  - !ruby/object:Gem::Version
18
- version: 0.1.3
60
+ version: 0.1.5
19
61
  type: :runtime
20
62
  prerelease: false
21
63
  version_requirements: !ruby/object:Gem::Requirement
22
64
  requirements:
23
65
  - - ">="
24
66
  - !ruby/object:Gem::Version
25
- version: 0.1.3
67
+ version: 0.1.5
26
68
  description: vLLM provider integration for the LegionIO LLM routing framework.
27
69
  email:
28
70
  - matthewdiverson@gmail.com
@@ -42,6 +84,10 @@ files:
42
84
  - lex-llm-vllm.gemspec
43
85
  - lib/legion/extensions/llm/vllm.rb
44
86
  - lib/legion/extensions/llm/vllm/provider.rb
87
+ - lib/legion/extensions/llm/vllm/registry_event_builder.rb
88
+ - lib/legion/extensions/llm/vllm/registry_publisher.rb
89
+ - lib/legion/extensions/llm/vllm/transport/exchanges/llm_registry.rb
90
+ - lib/legion/extensions/llm/vllm/transport/messages/registry_event.rb
45
91
  - lib/legion/extensions/llm/vllm/version.rb
46
92
  homepage: https://github.com/LegionIO/lex-llm-vllm
47
93
  licenses: