Module: Legion::LLM::API::Native::Models

Extended by:
Legion::Logging::Helper
Defined in:
lib/legion/llm/api/native/models.rb

Constant Summary collapse

AUTO_ROUTING_MODEL_ID = 
'legionio'
AUTO_ROUTING_MODEL_DISPLAY = 
'LegionIO'
AUTO_ROUTING_OFFERING_ID = 
'legionio:auto:inference:legionio'
AUTO_ROUTING_CAPABILITIES = 
%w[auto_routing chat completion json_schema tools].freeze

Class Method Summary collapse

Class Method Details

.auto_routing_alias_offering ⇒ Object 



268
269
270
271
272
273
274
275
276
277
 
# File 'lib/legion/llm/api/native/models.rb', line 268

def self.auto_routing_alias_offering
  base = auto_routing_offering
  base.merge(
    id:                    'legionio:auto:inference:auto',
    offering_id:           'legionio:auto:inference:auto',
    model:                 'auto',
    display_name:          'LegionIO (auto)',
    canonical_model_alias: 'auto'
  )
end

.auto_routing_model?(model) ⇒ Boolean

Returns:

  • (Boolean) 


328
329
330
331
332
333
334
335
336
337
 
# File 'lib/legion/llm/api/native/models.rb', line 328

def self.auto_routing_model?(model)
  m = model.to_s.strip.downcase
  # SSOT v4: read auto_routing_model_aliases from the canonical
  # [:llm][:router] home (default in settings/router.rb) so the catalog
  # view agrees with ingress (request.rb) and the Router (filter.rb).
  configured = Legion::Settings[:llm][:router][:auto_routing_model_aliases]
  aliases = Array(configured).map { |entry| entry.to_s.strip.downcase }.reject(&:empty?)
  aliases = [AUTO_ROUTING_MODEL_ID] if aliases.empty?
  aliases.include?(m)
end

.auto_routing_offering ⇒ Object 



279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
 
# File 'lib/legion/llm/api/native/models.rb', line 279

def self.auto_routing_offering
  ctx = Legion::Settings[:llm][:context_window]
  max_out = Legion::Settings[:llm][:max_output_tokens]
  {
    id:                    AUTO_ROUTING_OFFERING_ID,
    offering_id:           AUTO_ROUTING_OFFERING_ID,
    model:                 AUTO_ROUTING_MODEL_ID,
    display_name:          AUTO_ROUTING_MODEL_DISPLAY,
    model_family:          'legionio',
    canonical_model_alias: AUTO_ROUTING_MODEL_ID,
    type:                  :inference,
    provider_family:       'legionio',
    provider_instance:     'auto',
    instance_id:           'auto',
    tier:                  :auto,
    transport:             :internal,
    enabled:               true,
    capabilities:          AUTO_ROUTING_CAPABILITIES,
    limits:                { context_window: ctx, max_output_tokens: max_out },
    health:                { circuit_state: 'available' },
    metadata:              { auto_route: true, placeholder: true, display_name: AUTO_ROUTING_MODEL_DISPLAY },
    routing_metadata:      { strategy: 'auto' },
    source:                'static'
  }
end

.auto_routing_offering_matches?(filters) ⇒ Boolean

Returns:

  • (Boolean) 


305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
 
# File 'lib/legion/llm/api/native/models.rb', line 305

def self.auto_routing_offering_matches?(filters)
  normalized = request_filters(filters)
  type = normalized[:type]
  return false if type && !type.to_s.empty? && type.to_s != 'inference' && type.to_s != 'chat'

  provider = normalized[:provider]
  return false if provider && !provider.to_s.empty? && !%w[legionio auto].include?(provider.to_s.downcase)

  instance = normalized[:instance_id]
  return false if instance && !instance.to_s.empty? && !%w[auto legionio].include?(instance.to_s.downcase)

  model = normalized[:model] || normalized[:offering_id]
  return false if model && !model.to_s.empty? && !auto_routing_model?(model) && model.to_s != AUTO_ROUTING_OFFERING_ID

  family = normalized[:model_family]
  return false if family && !family.to_s.empty? && family.to_s.downcase != 'legionio'

  capability = normalized[:capability]
  return false if capability && !AUTO_ROUTING_CAPABILITIES.include?(capability.to_s)

  true
end

.health_display(availability) ⇒ Object

The 4-key legacy display shape (the same shape the provider actors write to the settings health hash), derived from the snapshot AvailabilityFact. Display only — selection reads the AvailabilityFact itself, never this projection.



138
139
140
141
142
143
144
145
146
147
148
149
150
 
# File 'lib/legion/llm/api/native/models.rb', line 138

def self.health_display(availability)
  available = availability&.state == :available
  {
    circuit_state: if available
                     :closed
                   else
                     (availability ? :open : :half_open)
                   end,
    denied:        false,
    available:     available,
    adjustment:    available ? 0 : -50
  }
end

.lane_entries(filters = {}) ⇒ Object

── v0.15.2 lane-hash projection (the single shared shaping) ──────────

Projects registry lanes onto the v0.15.2 lane-hash shape the models routes have always returned:

id, tier, provider_family, instance_id, model, canonical_model_alias,
type, capabilities, limits, enabled, cost, lane_weight,
health{circuit_state,denied,available,adjustment}

id/offering_id are the 5-tuple lane id (tier:provider_family: instance_id:type:model — composed only by Inventory::Identity.compose_lane_id at the writer). Every models route (native tree, namespaced tree, cross-namespace /api/llm/providers/:name/models, flat /v1/models tree) reads through lane_entries — there is no second projection.



95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
 
# File 'lib/legion/llm/api/native/models.rb', line 95

def self.lane_entries(filters = {})
  snapshot = Legion::LLM::Inventory.snapshot
  normalized = normalize_filter_hash(filters)
  inst_by_key = {}
  snapshot.each_instance { |inst| inst_by_key[inst.instance_key] = inst }

  snapshot.each_lane.filter_map do |lane|
    next unless lane_matches_filters?(lane, normalized)
    # Compliance-by-absence: a policy-denied model never
    # appears in the API surface (the registry publishes the full
    # catalog; the display applies the same policy the router does).
    next unless Legion::LLM::API::Native::Offerings.policy_permits?(lane)

    lane_entry(lane, inst_by_key[lane.instance_key]&.availability)
  end
end

.lane_entry(lane, availability) ⇒ Object 



112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
 
# File 'lib/legion/llm/api/native/models.rb', line 112

def self.lane_entry(lane, availability)
  {
    id:                    lane.lane_id,
    offering_id:           lane.lane_id,
    model:                 lane.model,
    provider_family:       lane.provider_family.to_s,
    provider_instance:     lane.instance_id.to_s,
    instance_id:           lane.instance_id.to_s,
    tier:                  lane.tier.to_s,
    type:                  lane_type(lane).to_s,
    canonical_model_alias: lane.metadata[:canonical_model_alias],
    model_family:          lane.metadata[:model_family],
    capabilities:          Legion::LLM::API::Native::Tiers.lane_capabilities(lane),
    limits:                Legion::LLM::API::Native::Tiers.lane_limits(lane),
    enabled:               true,
    cost:                  {},
    lane_weight:           lane_weight_for(lane, availability),
    health:                health_display(availability),
    metadata:              lane.metadata
  }
end

.lane_matches_filters?(lane, filters) ⇒ Boolean

Returns:

  • (Boolean) 


166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
 
# File 'lib/legion/llm/api/native/models.rb', line 166

def self.lane_matches_filters?(lane, filters)
  provider = filters[:provider]
  return false if provider && !provider.to_s.empty? && lane.provider_family.to_s != provider.to_s

  instance = filters[:instance_id]
  return false if instance && !instance.to_s.empty? && lane.instance_id.to_s != instance.to_s

  type = filters[:type]
  return false if type && !type.to_s.empty? && normalized_type(type) != lane_type(lane)

  model = filters[:model]
  return false if model && !model.to_s.empty? && lane.model.to_s != model.to_s

  offering_id = filters[:offering_id]
  return false if offering_id && !offering_id.to_s.empty? && lane.lane_id != offering_id.to_s

  family = filters[:model_family]
  return false if family && !family.to_s.empty? && lane.metadata[:model_family].to_s != family.to_s

  capability = filters[:capability]
  if capability && !capability.to_s.empty?
    caps = Legion::LLM::API::Native::Tiers.lane_capabilities(lane)
                                          .map { |c| Legion::Extensions::Llm::Capabilities.canonical(c) }
    return false unless caps.include?(Legion::Extensions::Llm::Capabilities.canonical(capability.to_s))
  end

  true
end

.lane_type(lane) ⇒ Object 



162
163
164
 
# File 'lib/legion/llm/api/native/models.rb', line 162

def self.lane_type(lane)
  Legion::Extensions::Llm::Taxonomies.lane_type_for(operation: lane.operation)
end

.lane_weight_for(lane, availability) ⇒ Object

v0.15.2 display law: lane_weight = tier_w × provider_w × instance_w × model_w (the write-time base_weight stored on the lane) × health multiplier. Available → ×1; unavailable (open circuit) → ×−1. Display only — the router reads the stored scalar and the AvailabilityFact, never this value.



157
158
159
160
 
# File 'lib/legion/llm/api/native/models.rb', line 157

def self.lane_weight_for(lane, availability)
  multiplier = availability&.state == :available ? 1 : -1
  lane.base_weight * multiplier
end

.model_summaries(offerings) ⇒ Object 



223
224
225
226
227
228
 
# File 'lib/legion/llm/api/native/models.rb', line 223

def self.model_summaries(offerings)
  summaries = offerings.group_by { |offering| offering[:model] }.map do |model, rows|
    summarize_model(model, rows)
  end
  summaries.sort_by { |model| [auto_routing_model?(model[:id]) ? 0 : 1, model[:id]] }
end

.normalize_filter_hash(filters) ⇒ Object 



207
208
209
 
# File 'lib/legion/llm/api/native/models.rb', line 207

def self.normalize_filter_hash(filters)
  filters.transform_keys(&:to_sym).compact
end

.normalized_type(value) ⇒ Object

Request filter spellings for the coarse lane type: the embedding spellings collapse to :embedding, 'chat' to :inference (the v0.15.2 request vocabulary), everything else passes through as a Taxonomies::TYPES symbol.



199
200
201
202
203
204
205
 
# File 'lib/legion/llm/api/native/models.rb', line 199

def self.normalized_type(value)
  case value.to_s
  when 'embedding', 'embeddings', 'embed' then :embedding
  when 'chat' then :inference
  else value.to_sym
  end
end

.registered(app) ⇒ Object 



20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
 
# File 'lib/legion/llm/api/native/models.rb', line 20

def self.registered(app)
  log.debug('[llm][api][models] registering model inventory routes')

  app.get '/api/llm/models' do
    log.debug('[llm][api][models] action=list_models')
    require_llm!

    filters = Legion::LLM::API::Native::Models.request_filters(params)
    offerings = Legion::LLM::API::Native::Models.lane_entries(filters)
    offerings = Legion::LLM::API::Native::Models.with_auto_routing_offering(offerings, filters)

    json_response({
                    models:    Legion::LLM::API::Native::Models.model_summaries(offerings),
                    offerings: offerings,
                    summary:   Legion::LLM::API::Native::Models.summary(offerings)
                  })
  rescue StandardError => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.models.list')
    json_error('model_inventory_error', e.message, status_code: 500)
  end

  app.get '/api/llm/models/:id' do
    model_id = params[:id]
    log.debug("[llm][api][models] action=get_model id=#{model_id}")
    require_llm!

    filters = { model: model_id }
    offerings = Legion::LLM::API::Native::Models.lane_entries(filters)
    offerings = Legion::LLM::API::Native::Models.with_auto_routing_offering(offerings, filters)
    halt json_error('model_not_found', "Model '#{model_id}' not found", status_code: 404) unless offerings.any?

    json_response({
                    model:     Legion::LLM::API::Native::Models.summarize_model(model_id, offerings),
                    offerings: offerings
                  })
  rescue StandardError => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.models.get')
    json_error('model_inventory_error', e.message, status_code: 500)
  end

  app.get '/api/llm/providers/:name/models' do
    provider = params[:name]
    log.debug("[llm][api][models] action=list_provider_models provider=#{provider}")
    require_llm!

    filters = Legion::LLM::API::Native::Models.request_filters(params).merge(provider: provider)
    offerings = Legion::LLM::API::Native::Models.lane_entries(filters)

    json_response({
                    provider:  provider,
                    models:    Legion::LLM::API::Native::Models.model_summaries(offerings),
                    offerings: offerings,
                    summary:   Legion::LLM::API::Native::Models.summary(offerings)
                  })
  rescue StandardError => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.models.provider')
    json_error('model_inventory_error', e.message, status_code: 500)
  end

  log.debug('[llm][api][models] model inventory routes registered')
end

.request_filters(params) ⇒ Object 



211
212
213
214
215
216
217
218
219
220
221
 
# File 'lib/legion/llm/api/native/models.rb', line 211

def self.request_filters(params)
  {
    provider:     params[:provider],
    instance_id:  params[:instance_id] || params[:instance],
    type:         params[:type] || params[:purpose],
    model:        params[:model],
    offering_id:  params[:offering_id],
    model_family: params[:model_family],
    capability:   params[:capability]
  }
end

.summarize_model(model, offerings) ⇒ Object 



230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
 
# File 'lib/legion/llm/api/native/models.rb', line 230

def self.summarize_model(model, offerings)
  summary = {
    id:             model.to_s,
    types:          offerings.map { |offering| offering[:type].to_s }.uniq.sort,
    providers:      offerings.map { |offering| offering[:provider_family] }.uniq.sort,
    model_families: offerings.filter_map { |offering| offering[:model_family] }.uniq.sort,
    offering_ids:   offerings.filter_map { |offering| offering[:offering_id] }.uniq.sort,
    instances:      offerings.map { |offering| offering[:instance_id] }.uniq.sort,
    capabilities:   offerings.flat_map { |offering| offering[:capabilities] }.uniq.sort,
    max_context:    offerings.filter_map { |offering| offering.dig(:limits, :context_window) }.max,
    enabled:        offerings.any? { |offering| offering[:enabled] != false }
  }
  if auto_routing_model?(model)
    first_display = offerings.filter_map { |o| o[:display_name] }.first
    summary[:display_name] = first_display || AUTO_ROUTING_MODEL_DISPLAY
    summary[:auto_route] = true
    summary[:default] = model.to_s == AUTO_ROUTING_MODEL_ID
  end
  summary
end

.summary(offerings) ⇒ Object 



251
252
253
254
255
256
257
258
259
 
# File 'lib/legion/llm/api/native/models.rb', line 251

def self.summary(offerings)
  {
    total_offerings: offerings.size,
    models:          offerings.map { |offering| offering[:model] }.uniq.size,
    providers:       offerings.map { |offering| offering[:provider_family] }.uniq.size,
    by_type:         offerings.group_by { |offering| offering[:type].to_s }
                              .transform_values(&:size)
  }
end

.with_auto_routing_offering(offerings, filters = {}) ⇒ Object 



261
262
263
264
265
266
 
# File 'lib/legion/llm/api/native/models.rb', line 261

def self.with_auto_routing_offering(offerings, filters = {})
  return offerings unless auto_routing_offering_matches?(filters)
  return offerings if offerings.any? { |offering| auto_routing_model?(offering[:model]) }

  [auto_routing_offering, auto_routing_alias_offering, *offerings]
end