Module: Legion::LLM::API::OpenAI::ChatCompletions

Extended by:
Legion::Logging::Helper
Defined in:
lib/legion/llm/api/openai/chat_completions.rb

Class Method Summary collapse

Class Method Details

.append_usage_stats(done_chunk, pipeline_response, include_reasoning) ⇒ Object

Append usage stats to the done chunk when reasoning is enabled.



257
258
259
260
261
262
263
264
265
266
267
268
269
 
# File 'lib/legion/llm/api/openai/chat_completions.rb', line 257

def self.append_usage_stats(done_chunk, pipeline_response, include_reasoning)
  return unless include_reasoning

  tokens = pipeline_response.tokens || {}
  oai = Legion::LLM::API::Translators::OpenAIResponse
  input_count = oai.extract_token_count(tokens, :input).to_i
  output_count = oai.extract_token_count(tokens, :output).to_i
  done_chunk[:usage] = {
    prompt_tokens:     input_count,
    completion_tokens: output_count,
    total_tokens:      input_count + output_count
  }
end

.build_handler ⇒ Object

rubocop:disable Metrics/MethodLength,Metrics/AbcSize



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
 
# File 'lib/legion/llm/api/openai/chat_completions.rb', line 25

def self.build_handler # rubocop:disable Metrics/MethodLength,Metrics/AbcSize
  proc do # rubocop:disable Metrics/BlockLength
    require_llm!
    body = parse_request_body

    unless body[:messages].is_a?(Array) && !body[:messages].empty?
      halt 400, { 'Content-Type' => 'application/json' },
           Legion::JSON.dump({ error: { message: 'messages is required and must be a non-empty array',
                                        type: 'invalid_request_error', param: 'messages', code: nil } })
    end

    request_id = body[:request_id] || SecureRandom.uuid
    normalized = Legion::LLM::API::Translators::OpenAIRequest.normalize(body)
    model = normalized[:model] || Legion::LLM::Settings.value(:default_model) || 'default'
    streaming = normalized[:stream] == true
    include_reasoning = body[:include_reasoning] == true || body[:include_thinking] == true

    # ── Extended fields + pipeline request (parity with native) ─────────
    ext = Legion::LLM::API::OpenAI::ChatCompletions.extract_extended_fields(body, env)

    log.info('[llm][api][openai][chat_completions] action=accepted ' \
             "request_id=#{request_id} model=#{model} stream=#{streaming} " \
             "conversation_id=#{ext[:conversation_id] || 'none'} tier=#{ext[:tier] || 'auto'}")

    tool_declarations = Legion::LLM::API::OpenAI::ChatCompletions.build_openai_tool_classes(normalized[:tools])

    # ── Gaia ingest (mirrors native endpoint pre-pipeline awareness) ────
    Legion::LLM::API::OpenAI::ChatCompletions.gaia_ingest(
      body[:messages], request_id, identity_canonical_name(env)
    )

    effective_caller = build_server_caller(
      source: 'openai_compat', path: request.path, env: env,
      caller_context: ext[:caller_context]
    )

    inference_request = Legion::LLM::API::OpenAI::ChatCompletions.build_inference_request(
      request_id: request_id, normalized: normalized, model: model,
      tool_declarations: tool_declarations, caller: effective_caller,
      streaming: streaming, ext: ext
    )

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

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

      stream do |out| # rubocop:disable Metrics/BlockLength
        pipeline_response = executor.call_stream do |chunk|
          Legion::LLM::API::OpenAI::ChatCompletions.emit_reasoning_delta(
            out, chunk, model, request_id, include_reasoning
          )
          text = chunk.respond_to?(:content) ? chunk.content.to_s : chunk.to_s
          next if text.empty?

          chunk_obj = Legion::LLM::API::Translators::OpenAIResponse.format_stream_chunk(
            text, model: model, request_id: request_id
          )
          out << "data: #{Legion::JSON.dump(chunk_obj)}\n\n"
        end

        routing = pipeline_response.routing || {}
        final_model = (routing[:model] || routing['model'] || model).to_s
        tool_calls = Legion::LLM::API::Translators::OpenAIResponse.build_tool_calls(pipeline_response)
        tool_calls.each_with_index do |tool_call, index|
          tc_chunk = Legion::LLM::API::Translators::OpenAIResponse.format_stream_tool_call_chunk(
            tool_call, model: final_model, request_id: request_id, index: index
          )
          out << "data: #{Legion::JSON.dump(tc_chunk)}\n\n"
        end
        done_chunk = Legion::LLM::API::Translators::OpenAIResponse.format_stream_chunk(
          nil,
          model:         final_model,
          request_id:    request_id,
          finish_reason: tool_calls.empty? ? 'stop' : 'tool_calls',
          usage:         {
            prompt_tokens:     Legion::LLM::API::Translators::OpenAIResponse.extract_token_count(pipeline_response.tokens, :input),
            completion_tokens: Legion::LLM::API::Translators::OpenAIResponse.extract_token_count(pipeline_response.tokens, :output),
            total_tokens:      Legion::LLM::API::Translators::OpenAIResponse.extract_token_count(pipeline_response.tokens, :input).to_i +
                               Legion::LLM::API::Translators::OpenAIResponse.extract_token_count(pipeline_response.tokens, :output).to_i
          }
        )
        Legion::LLM::API::OpenAI::ChatCompletions.append_usage_stats(
          done_chunk, pipeline_response, include_reasoning
        )
        out << "data: #{Legion::JSON.dump(done_chunk)}\n\n"
        out << "data: [DONE]\n\n"
        log.info('[llm][api][openai][chat_completions] action=stream_complete ' \
                 "request_id=#{request_id} model=#{final_model}")
      rescue StandardError => e
        handle_exception(e, level: :error, handled: false,
                         operation: 'llm.api.openai.chat_completions.stream',
                         request_id: request_id)
        out << "data: #{Legion::JSON.dump({ error: { message: e.message, type: 'server_error' } })}\n\n"
        out << "data: [DONE]\n\n"
      end
    else
      pipeline_response = executor.call
      response_body = Legion::LLM::API::Translators::OpenAIResponse.format_chat_completion(
        pipeline_response, model: model, request_id: request_id,
        include_reasoning: include_reasoning
      )

      log.info("[llm][api][openai][chat_completions] action=complete request_id=#{request_id} model=#{response_body[:model]}")
      content_type :json
      status 200
      Legion::JSON.dump(response_body)
    end
  rescue Legion::LLM::AuthError => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.openai.chat_completions.auth')
    halt 401, { 'Content-Type' => 'application/json' },
         Legion::JSON.dump({ error: { message: e.message, type: 'authentication_error' } })
  rescue Legion::LLM::RateLimitError => e
    handle_exception(e, level: :warn, handled: true, operation: 'llm.api.openai.chat_completions.rate_limit')
    halt 429, { 'Content-Type' => 'application/json' },
         Legion::JSON.dump({ error: { message: e.message, type: 'requests', code: 'rate_limit_exceeded' } })
  rescue Legion::LLM::ProviderDown, Legion::LLM::ProviderError => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.openai.chat_completions.provider')
    halt 502, { 'Content-Type' => 'application/json' },
         Legion::JSON.dump({ error: { message: e.message, type: 'server_error' } })
  rescue StandardError => e
    handle_exception(e, level: :error, handled: false, operation: 'llm.api.openai.chat_completions')
    halt 500, { 'Content-Type' => 'application/json' },
         Legion::JSON.dump({ error: { message: e.message, type: 'server_error' } })
  end
end

.build_inference_request(request_id:, normalized:, model:, tool_declarations:, caller:, streaming:, ext:) ⇒ Object

Build the Inference::Request with full pipeline field set.



177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
 
# File 'lib/legion/llm/api/openai/chat_completions.rb', line 177

def self.build_inference_request(request_id:, normalized:, model:, tool_declarations:, caller:, streaming:, ext:)
  extra = {}
  extra[:tier] = ext[:tier].to_sym if ext[:tier]
  extra[:cwd] = ext[:cwd] if ext[:cwd]

  metadata = { requested_tools: ext[:requested_tools] }
  metadata[:client_tool_passthrough] = ext[:client_tool_passthrough] unless ext[:client_tool_passthrough].nil?
  metadata[:client_tool_request_count] = normalized[:tools]&.size if normalized[:tools]&.any?

  Legion::LLM::Inference::Request.build(
    id:              request_id,
    messages:        normalized[:messages],
    system:          normalized[:system],
    routing:         { provider: ext[:provider], model: model, instance: ext[:instance] }.compact,
    tools:           tool_declarations,
    caller:          caller,
    conversation_id: ext[:conversation_id],
    metadata:        metadata.compact,
    stream:          streaming,
    cache:           { strategy: :default, cacheable: true },
    extra:           extra.empty? ? {} : extra
  )
end

.build_openai_tool_classes(tools) ⇒ Object 



201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
 
# File 'lib/legion/llm/api/openai/chat_completions.rb', line 201

def self.build_openai_tool_classes(tools)
  return [] if tools.nil? || !tools.is_a?(Array) || tools.empty?

  tools.filter_map do |tool|
    t = nil
    t = tool.respond_to?(:transform_keys) ? tool.transform_keys(&:to_sym) : tool
    next unless t[:name].to_s.length.positive?

    Legion::LLM::Types::ToolDefinition.build(
      name:        t[:name].to_s,
      description: t[:description].to_s,
      parameters:  t[:parameters] || {},
      source:      { type: :client, executable: true }
    )
  rescue StandardError => e
    tool_name = t.is_a?(Hash) ? t[:name] : nil
    handle_exception(e, level: :warn, handled: true, operation: "llm.api.openai.build_tool.#{tool_name || 'unknown'}")
    nil
  end
end

.emit_reasoning_delta(out, chunk, model, request_id, include_reasoning) ⇒ Object

Emit a reasoning_content delta chunk if the streaming chunk contains thinking.



243
244
245
246
247
248
249
250
251
252
253
254
 
# File 'lib/legion/llm/api/openai/chat_completions.rb', line 243

def self.emit_reasoning_delta(out, chunk, model, request_id, include_reasoning)
  return unless include_reasoning && chunk.respond_to?(:thinking)

  thinking_text = extract_chunk_text(chunk.thinking)
  return if thinking_text.empty?

  reasoning_chunk = Legion::LLM::API::Translators::OpenAIResponse.format_stream_delta_chunk(
    { reasoning_content: thinking_text },
    model: model, request_id: request_id
  )
  out << "data: #{Legion::JSON.dump(reasoning_chunk)}\n\n"
end

.extract_chunk_text(value) ⇒ Object

Extract text content from a thinking chunk value. Handles the various shapes the chunk.thinking field can take:

- Hash with :content or :text key
- Object with .content or .text method
- Raw string



227
228
229
230
231
232
233
234
235
236
237
238
239
240
 
# File 'lib/legion/llm/api/openai/chat_completions.rb', line 227

def self.extract_chunk_text(value)
  return '' if value.nil?
  return value.to_s if value.is_a?(String)

  if value.is_a?(Hash)
    text = value[:content] || value['content'] || value[:text] || value['text']
    return text.to_s if text
  end

  return value.content.to_s if value.respond_to?(:content) && value.content
  return value.text.to_s if value.respond_to?(:text) && value.text

  value.to_s
end

.extract_extended_fields(body, env) ⇒ Object

Extract extended pipeline fields from body + X-Legion-* headers. Headers take precedence for scalar values; body for complex objects.



157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
 
# File 'lib/legion/llm/api/openai/chat_completions.rb', line 157

def self.extract_extended_fields(body, env)
  ctp = if env.key?('HTTP_X_LEGION_CLIENT_TOOL_PASSTHROUGH')
          env['HTTP_X_LEGION_CLIENT_TOOL_PASSTHROUGH'] == 'true'
        elsif [true, false].include?(body[:client_tool_passthrough])
          body[:client_tool_passthrough]
        end

  {
    conversation_id:         env['HTTP_X_LEGION_CONVERSATION_ID'] || body[:conversation_id],
    provider:                env['HTTP_X_LEGION_PROVIDER'] || body[:provider],
    tier:                    env['HTTP_X_LEGION_TIER'] || body[:tier],
    instance:                env['HTTP_X_LEGION_INSTANCE'] || body[:instance],
    cwd:                     env['HTTP_X_LEGION_CWD'] || body[:cwd],
    requested_tools:         body[:requested_tools] || [],
    client_tool_passthrough: ctp,
    caller_context:          body[:caller]
  }
end

.gaia_ingest(messages, request_id, caller_identity) ⇒ Object

Pre-pipeline Gaia ingest — mirrors the native endpoint’s awareness update. Feeds the latest user prompt into Gaia so advisory/context is fresh.



273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
 
# File 'lib/legion/llm/api/openai/chat_completions.rb', line 273

def self.gaia_ingest(messages, request_id, caller_identity)
  return unless defined?(Legion::Gaia) && Legion::Gaia.respond_to?(:started?) && Legion::Gaia.started?

  last_user = Array(messages).select { |m| (m[:role] || m['role']).to_s == 'user' }.last
  prompt = (last_user || {})[:content] || (last_user || {})['content'] || ''
  return if prompt.to_s.empty?

  frame = Legion::Gaia::InputFrame.new(
    content:      prompt.to_s,
    channel_id:   :api,
    content_type: :text,
    auth_context: { identity: caller_identity },
    metadata:     { source_type: :human_direct, salience: 0.9 }
  )
  Legion::Gaia.ingest(frame)
  log.debug("[llm][api][openai][chat_completions] action=gaia_ingest request_id=#{request_id}")
rescue StandardError => e
  log.warn("[llm][api][openai][chat_completions] gaia_ingest failed: #{e.message}")
end

.registered(app) ⇒ Object 



14
15
16
17
18
19
20
21
22
23
 
# File 'lib/legion/llm/api/openai/chat_completions.rb', line 14

def self.registered(app)
  log.debug('[llm][api][openai][chat_completions] registering POST /v1/chat/completions + /api/llm/inference/v1/chat/completions')

  handler = build_handler

  app.post('/v1/chat/completions') { instance_exec(&handler) }
  app.post('/api/llm/inference/v1/chat/completions') { instance_exec(&handler) }

  log.debug('[llm][api][openai][chat_completions] routes registered')
end