Class: Legion::Extensions::Llm::Vllm::Provider

Inherits:
Provider
  • Object
show all
Includes:
Provider::OpenAICompatible, Logging::Helper
Defined in:
lib/legion/extensions/llm/vllm/provider.rb

Overview

vLLM provider implementation for the Legion::Extensions::Llm base provider contract.

Defined Under Namespace

Modules: Capabilities

Class Method Summary collapse

Instance Method Summary collapse

Class Method Details

.capabilitiesObject



20
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 20

def capabilities = Capabilities

.configuration_optionsObject



18
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 18

def configuration_options = %i[vllm_api_base vllm_api_key]

.configuration_requirementsObject



19
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 19

def configuration_requirements = []

.local?Boolean

Returns:

  • (Boolean)


17
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 17

def local? = false

.registry_publisherObject



22
23
24
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 22

def registry_publisher
  Vllm.registry_publisher
end

.slugObject



16
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 16

def slug = 'vllm'

Instance Method Details

#api_baseObject



53
54
55
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 53

def api_base
  normalize_url(config.vllm_api_base || 'localhost:8000')
end

#headersObject



57
58
59
60
61
62
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 57

def headers
  token = config.vllm_api_key
  return {} if token.nil? || token.to_s.empty?

  { 'Authorization' => "Bearer #{token}" }
end

#healthObject



71
72
73
74
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 71

def health
  log.info { "checking health at #{api_base}#{health_url}" }
  connection.get(health_url).body
end

#health_urlObject



64
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 64

def health_url = '/health'

#list_modelsObject



83
84
85
86
87
88
89
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 83

def list_models
  log.info { "discovering models from #{api_base}#{models_url}" }
  super.tap do |models|
    log.info { "discovered #{models.size} model(s) from vLLM" }
    self.class.registry_publisher.publish_models_async(models, readiness: readiness(live: false))
  end
end

#readiness(live: false) ⇒ Object



76
77
78
79
80
81
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 76

def readiness(live: false)
  log.info { "checking readiness live=#{live} at #{api_base}" }
  super.tap do ||
    self.class.registry_publisher.publish_readiness_async() if live
  end
end

#reset_mm_cacheObject



100
101
102
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 100

def reset_mm_cache
  connection.post(reset_mm_cache_url, {}).body
end

#reset_mm_cache_urlObject



67
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 67

def reset_mm_cache_url = '/reset_mm_cache'

#reset_prefix_cache(reset_running_requests: nil, reset_external: nil) ⇒ Object



96
97
98
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 96

def reset_prefix_cache(reset_running_requests: nil, reset_external: nil)
  connection.post(with_query(reset_prefix_cache_url, reset_running_requests:, reset_external:), {}).body
end

#reset_prefix_cache_urlObject



66
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 66

def reset_prefix_cache_url = '/reset_prefix_cache'

#settingsObject



49
50
51
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 49

def settings
  Vllm.default_settings
end

#sleep(level: 1) ⇒ Object



104
105
106
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 104

def sleep(level: 1)
  connection.post(with_query(sleep_url, level:), {}).body
end

#sleep_urlObject



68
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 68

def sleep_url = '/sleep'

#stream_usage_supported?Boolean

Returns:

  • (Boolean)


47
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 47

def stream_usage_supported? = true

#versionObject



91
92
93
94
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 91

def version
  log.info { "fetching version from #{api_base}#{version_url}" }
  connection.get(version_url).body
end

#version_urlObject



65
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 65

def version_url = '/version'

#wake_up(tags: nil) ⇒ Object



108
109
110
111
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 108

def wake_up(tags: nil)
  query = Array(tags).map { |tag| ['tags', tag] }
  connection.post(with_query(wake_up_url, query), {}).body
end

#wake_up_urlObject



69
# File 'lib/legion/extensions/llm/vllm/provider.rb', line 69

def wake_up_url = '/wake_up'