Module: Legion::LLM::API::Native::Inference

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

Class Method Summary collapse

Class Method Details

.registered(app) ⇒ Object

rubocop:disable Metrics/MethodLength,Metrics/AbcSize,Metrics/CyclomaticComplexity,Metrics/PerceivedComplexity



13
14
15
16
17
18
19
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
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
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
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
 
# File 'lib/legion/llm/api/native/inference.rb', line 13

def self.registered(app) # rubocop:disable Metrics/MethodLength,Metrics/AbcSize,Metrics/CyclomaticComplexity,Metrics/PerceivedComplexity
  log.debug('[llm][api][inference] registering POST /api/llm/inference')

  app.post '/api/llm/inference' do # rubocop:disable Metrics/BlockLength
    require_llm!
    body = parse_request_body
    validate_required!(body, :messages)

    messages        = body[:messages]
    raw_tools       = body[:tools]
    requested_tools = body[:requested_tools] || []
    model           = body[:model]
    provider        = body[:provider]
    caller_context  = body[:caller]
    conversation_id = body[:conversation_id]
    request_id      = body[:request_id] || SecureRandom.uuid

    unless messages.is_a?(Array)
      halt 400, { 'Content-Type' => 'application/json' },
           Legion::JSON.dump({ error: { code: 'invalid_messages', message: 'messages must be an array' } })
    end

    validate_messages!(messages)

    unless raw_tools.nil? || raw_tools.is_a?(Array)
      halt 400, { 'Content-Type' => 'application/json' },
           Legion::JSON.dump({ error: { code: 'invalid_tools', message: 'tools must be an array' } })
    end

    tools = raw_tools || []
    validate_tools!(tools) unless tools.empty?

    caller_identity = resolve_caller_identity(env)
    last_user = messages.select { |m| (m[:role] || m['role']).to_s == 'user' }.last
    prompt    = (last_user || {})[:content] || (last_user || {})['content'] || ''

    route_t0 = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)

    if defined?(Legion::Gaia) && Legion::Gaia.respond_to?(:started?) && Legion::Gaia.started? && prompt.to_s.length.positive?
      begin
        gaia_t0 = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
        frame = Legion::Gaia::InputFrame.new(
          content:      prompt,
          channel_id:   :api,
          content_type: :text,
          auth_context: { identity: caller_identity },
          metadata:     { source_type: :human_direct, salience: 0.9 }
        )
        Legion::Gaia.ingest(frame)
        gaia_ms = ((::Process.clock_gettime(::Process::CLOCK_MONOTONIC) - gaia_t0) * 1000).round
        log.debug("[llm][api][inference] action=gaia_ingest duration_ms=#{gaia_ms} request_id=#{request_id}")
      rescue StandardError => e
        handle_exception(e, level: :warn, handled: true, operation: 'llm.api.inference.gaia_ingest', request_id: request_id)
      end
    end

    tool_declarations = tools.filter_map do |tool|
      ts = tool.respond_to?(:transform_keys) ? tool.transform_keys(&:to_sym) : tool
      build_client_tool_class(ts[:name].to_s, ts[:description].to_s, ts[:parameters] || ts[:input_schema])
    end

    log.debug("[llm][api][inference] action=tools_built client_tools=#{tool_declarations.size}")

    streaming = body[:stream] == true && request.preferred_type.to_s.include?('text/event-stream')
    normalized_caller = caller_context.respond_to?(:transform_keys) ? caller_context.transform_keys(&:to_sym) : {}
    safe_caller_fields = normalized_caller.slice(:context, :session_id, :trace_id)
    server_caller_fields = {
      source:       'api',
      path:         request.path,
      requested_by: resolve_requested_by(env, caller_identity)
    }
    effective_caller = server_caller_fields.merge(safe_caller_fields)
    caller_summary = [effective_caller[:source], effective_caller[:path]].compact.join(':')
    log.info(
      "[llm][api][inference] action=accepted request_id=#{request_id} " \
      "conversation_id=#{conversation_id || 'none'} caller=#{caller_summary} " \
      "messages=#{messages.size} client_tools=#{tools.size} requested_tools=#{Array(requested_tools).size} " \
      "requested_provider=#{provider || 'auto'} requested_model=#{model || 'auto'} stream=#{streaming}"
    )

    require 'legion/llm/inference/request' unless defined?(Legion::LLM::Inference::Request)
    require 'legion/llm/inference/executor' unless defined?(Legion::LLM::Inference::Executor)

    pipeline_request = Legion::LLM::Inference::Request.build(
      id:              request_id,
      messages:        messages,
      system:          body[:system],
      routing:         { provider: provider, model: model },
      tools:           tool_declarations,
      caller:          effective_caller,
      conversation_id: conversation_id,
      metadata:        { requested_tools: requested_tools },
      stream:          streaming,
      cache:           { strategy: :default, cacheable: true }
    )

    setup_ms = ((::Process.clock_gettime(::Process::CLOCK_MONOTONIC) - route_t0) * 1000).round
    log.debug("[llm][api][inference] action=pipeline_setup duration_ms=#{setup_ms} request_id=#{request_id}")

    executor = Legion::LLM::Inference::Executor.new(pipeline_request)

    if streaming
      content_type 'text/event-stream'
      headers 'Cache-Control'     => 'no-cache',
              'Connection'        => 'keep-alive',
              'X-Accel-Buffering' => 'no'

      # rubocop:disable Metrics/BlockLength
      stream do |out|
        full_text = +''

        executor.tool_event_handler = lambda { |event|
          log.info("[llm][api][inference] action=tool_event type=#{event[:type]} tool=#{event[:tool_name]} id=#{event[:tool_call_id]}")
          case event[:type]
          when :tool_call
            emit_sse_event(out, 'tool-call', {
                             toolCallId: event[:tool_call_id],
                             toolName:   event[:tool_name],
                             args:       event[:arguments],
                             timestamp:  Time.now.utc.iso8601
                           })
          when :tool_result
            emit_sse_event(out, 'tool-result', {
                             toolCallId: event[:tool_call_id],
                             toolName:   event[:tool_name],
                             result:     event[:result],
                             timestamp:  Time.now.utc.iso8601
                           })
          when :tool_error
            emit_sse_event(out, 'tool-error', {
                             toolCallId: event[:tool_call_id],
                             toolName:   event[:tool_name],
                             result:     event[:error],
                             status:     'error',
                             timestamp:  Time.now.utc.iso8601
                           })
          end
        }

        pipeline_response = executor.call_stream do |chunk|
          text = chunk.respond_to?(:content) ? chunk.content.to_s : chunk.to_s
          next if text.empty?

          full_text << text
          emit_sse_event(out, 'text-delta', { delta: text })
        end

        emit_timeline_tool_events(out, pipeline_response, skip_tool_results: !executor.tool_event_handler.nil?)

        enrichments = pipeline_response.enrichments
        emit_sse_event(out, 'enrichment', enrichments) if enrichments.is_a?(Hash) && !enrichments.empty?

        routing = pipeline_response.routing || {}
        tokens = pipeline_response.tokens || {}
        emit_sse_event(out, 'done', {
                         request_id:      request_id,
                         content:         full_text,
                         model:           (routing[:model] || routing['model']).to_s,
                         input_tokens:    token_value(tokens, :input),
                         output_tokens:   token_value(tokens, :output),
                         tool_calls:      extract_tool_calls(pipeline_response),
                         conversation_id: pipeline_response.conversation_id
                       })

        log.info(
          "[llm][api][inference] action=completed request_id=#{request_id} " \
          "conversation_id=#{pipeline_response.conversation_id || conversation_id || 'none'} " \
          "provider=#{routing[:provider] || routing['provider'] || 'unknown'} " \
          "model=#{routing[:model] || routing['model'] || 'unknown'} " \
          "tool_calls=#{extract_tool_calls(pipeline_response).size} " \
          "tool_executions=#{Array(pipeline_response.timeline).count { |event| event[:key].to_s.start_with?('tool:execute:') }} " \
          "stop_reason=#{pipeline_response.stop&.dig(:reason) || 'unknown'} stream=true"
        )
      rescue StandardError => e
        handle_exception(e, level: :error, handled: false, operation: 'llm.api.inference.stream', request_id: request_id)
        emit_sse_event(out, 'error', { code: 'stream_error', message: e.message })
      end
      # rubocop:enable Metrics/BlockLength
    else
      exec_t0 = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
      pipeline_response = executor.call
      exec_ms = ((::Process.clock_gettime(::Process::CLOCK_MONOTONIC) - exec_t0) * 1000).round
      log.debug("[llm][api][inference] action=executor_call duration_ms=#{exec_ms} request_id=#{request_id}")
      raw_msg = pipeline_response.message
      content = raw_msg.is_a?(Hash) ? (raw_msg[:content] || raw_msg['content']) : raw_msg.to_s
      routing = pipeline_response.routing || {}
      tokens = pipeline_response.tokens || {}
      tool_calls = extract_tool_calls(pipeline_response)

      log.info(
        "[llm][api][inference] action=completed request_id=#{request_id} " \
        "conversation_id=#{pipeline_response.conversation_id || conversation_id || 'none'} " \
        "provider=#{routing[:provider] || routing['provider'] || 'unknown'} " \
        "model=#{routing[:model] || routing['model'] || 'unknown'} " \
        "tool_calls=#{tool_calls.size} " \
        "tool_executions=#{Array(pipeline_response.timeline).count { |event| event[:key].to_s.start_with?('tool:execute:') }} " \
        "stop_reason=#{pipeline_response.stop&.dig(:reason) || 'unknown'} stream=false"
      )

      json_response({
                      request_id:      request_id,
                      content:         content,
                      tool_calls:      tool_calls,
                      stop_reason:     pipeline_response.stop&.dig(:reason)&.to_s,
                      model:           (routing[:model] || routing['model']).to_s,
                      input_tokens:    token_value(tokens, :input),
                      output_tokens:   token_value(tokens, :output),
                      conversation_id: pipeline_response.conversation_id
                    }, status_code: 200)
    end
  rescue Legion::LLM::AuthError => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.inference.auth', request_id: request_id)
    json_error('auth_error', e.message, status_code: 401)
  rescue Legion::LLM::RateLimitError => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.inference.rate_limit', request_id: request_id)
    json_error('rate_limit', e.message, status_code: 429)
  rescue Legion::LLM::TokenBudgetExceeded => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.inference.budget', request_id: request_id)
    json_error('token_budget_exceeded', e.message, status_code: 413)
  rescue Legion::LLM::ProviderDown, Legion::LLM::ProviderError => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.inference.provider', request_id: request_id)
    json_error('provider_error', e.message, status_code: 502)
  rescue StandardError => e
    handle_exception(e, level: :error, handled: false, operation: 'llm.api.inference', request_id: request_id)
    json_error('inference_error', e.message, status_code: 500)
  end

  log.debug('[llm][api][inference] POST /api/llm/inference registered')
end