Class: Legion::Extensions::Llm::AzureFoundry::Provider

Inherits:
Provider
  • Object
show all
Extended by:
ProviderClassMethods
Includes:
ProviderDispatchMethods, Provider::OpenAICompatible
Defined in:
lib/legion/extensions/llm/azure_foundry/provider.rb

Overview

Azure AI Foundry and Azure OpenAI hosted provider surface.

Offerings are produced ONLY by the discovery actor's writer path (OfferingDraft + Registry publication, 07 C1-C5); the base read path discover_offerings serves the activated inventory offerings from the Registry snapshot (07 C5). The legacy offering production path (and its capability-policy cascade) is gone.

Constant Summary collapse

DEFAULT_API_VERSION = 
'2024-05-01-preview'
MODEL_INFERENCE_SURFACE = 
:model_inference
OPENAI_V1_SURFACE = 
:openai_v1

Instance Method Summary collapse

Methods included from ProviderClassMethods

capabilities, configuration_options, configuration_requirements, default_tier, default_transport, slug

Methods included from ProviderDispatchMethods

#chat, #embed, #stream

Instance Method Details

#api_base ⇒ Object 



116
117
118
119
120
121
122
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 116

def api_base
  endpoint = config.azure_foundry_endpoint.to_s.sub(%r{/*\z}, '')
  return "#{endpoint}/openai/v1" if surface == OPENAI_V1_SURFACE && !endpoint.end_with?('/openai/v1')
  return endpoint.delete_suffix('/models') if surface == MODEL_INFERENCE_SURFACE

  endpoint
end

#chat_url ⇒ Object 



130
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 130

def chat_url = completion_url

#completion_url ⇒ Object 



129
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 129

def completion_url = path_for('chat/completions')

#embedding_url ⇒ Object 



133
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 133

def embedding_url(**) = path_for('embeddings')

#headers ⇒ Object 



124
125
126
127
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 124

def headers
  identity_headers.merge({ 'api-key' => config.azure_foundry_api_key,
                           'Authorization' => bearer_header }.compact)
end

#health(live: false) ⇒ Object 



136
137
138
139
140
141
142
143
144
145
146
147
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 136

def health(live: false)
  log.info { "checking health live=#{live} at #{api_base}" }
  baseline = health_baseline(live)
  return baseline.merge(checked: false) unless live

  response = connection.get(health_url)
  baseline.merge(checked: true, ready: true, status: 'healthy', raw: response.body)
rescue StandardError => e
  handle_exception(e, level: :warn, handled: true, operation: 'azure_foundry.health')
  baseline.merge(checked: true, ready: false, status: 'unhealthy',
                 error: e.class.name, message: e.message)
end

#health_url ⇒ Object 



134
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 134

def health_url = models_url

#model_id(model) ⇒ Object 



154
155
156
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 154

def model_id(model)
  model.respond_to?(:id) ? model.id : model.to_s
end

#models_url ⇒ Object 



132
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 132

def models_url = surface == MODEL_INFERENCE_SURFACE ? path_for('info') : path_for('models')

#readiness(live: false) ⇒ Object 



149
150
151
152
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 149

def readiness(live: false)
  log.info { "checking readiness live=#{live} at #{api_base}" }
  health(live: live).merge(local: false, remote: true, endpoints: endpoint_manifest)
end

#stream_url ⇒ Object 



131
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 131

def stream_url = completion_url

#stream_usage_supported? ⇒ Boolean

Returns:

  • (Boolean) 


114
 
# File 'lib/legion/extensions/llm/azure_foundry/provider.rb', line 114

def stream_usage_supported? = true