Module: Legion::LLM::API::Namespaces::OpenAI::Completions

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

Class Method Summary collapse

Class Method Details

.extract_token(tokens, key) ⇒ Object 



89
90
91
92
93
94
95
96
97
98
99
100
101
 
# File 'lib/legion/llm/api/namespaces/openai/completions.rb', line 89

def self.extract_token(tokens, key)
  return 0 if tokens.nil?

  if tokens.is_a?(Hash)
    v = tokens[key] || tokens[key.to_s]
    return v.to_i unless v.nil?
  end

  method_name = { input_tokens: :input_tokens, output_tokens: :output_tokens }[key]
  return tokens.public_send(method_name).to_i if method_name && tokens.respond_to?(method_name)

  0
end

.registered(app) ⇒ Object

rubocop:disable Metrics/MethodLength



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
 
# File 'lib/legion/llm/api/namespaces/openai/completions.rb', line 15

def self.registered(app) # rubocop:disable Metrics/MethodLength
  log.debug('[llm][api][namespaces][openai][completions] registering routes')

  # rubocop:disable Metrics/BlockLength
  app.post '/v1/completions' do
    require_llm!
    body   = parse_request_body
    prompt = body[:prompt]

    if prompt.nil? || (prompt.respond_to?(:empty?) && prompt.empty?)
      return openai_error('prompt is required', type: 'invalid_request_error',
                                              code: nil, status_code: 400)
    end

    request_id = SecureRandom.uuid
    model      = body[:model] || Legion::LLM::Settings.value(:default_model) || 'default'
    messages   = [{ role: 'user', content: prompt.to_s }]

    log.info("[llm][api][namespaces][openai][completions] action=accepted request_id=#{request_id} model=#{model}")

    inference_request = Legion::LLM::Inference::Request.build(
      id:       request_id,
      messages: messages,
      routing:  { model: model },
      tools:    [],
      caller:   build_server_caller(source: 'openai_completions', path: request.path, env: env),
      stream:   false,
      cache:    { strategy: :default, cacheable: true }
    )
    pipeline_response = Legion::LLM::Inference::Executor.new(inference_request).call

    routing        = pipeline_response.routing || {}
    tokens         = pipeline_response.tokens  || {}
    raw_msg        = pipeline_response.message
    text           = raw_msg.is_a?(Hash) ? (raw_msg[:content] || raw_msg['content']).to_s : raw_msg.to_s
    resolved_model = (routing[:model] || routing['model'] || model).to_s
    input_tokens  = Completions.extract_token(tokens, :input_tokens)
    output_tokens = Completions.extract_token(tokens, :output_tokens)

    log.info("[llm][api][namespaces][openai][completions] action=complete request_id=#{request_id} model=#{resolved_model}")
    content_type :json
    status 200
    Legion::JSON.dump({
                        id:      "cmpl-#{request_id.delete('-')}",
                        object:  'text_completion',
                        created: Time.now.to_i,
                        model:   resolved_model,
                        choices: [{ text: text, index: 0, finish_reason: 'stop' }],
                        usage:   {
                          prompt_tokens:     input_tokens,
                          completion_tokens: output_tokens,
                          total_tokens:      input_tokens.to_i + output_tokens.to_i
                        }
                      })
  rescue Legion::LLM::AuthError => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.namespaces.openai.completions.auth')
    openai_error(e.message, type: 'authentication_error', status_code: 401)
  rescue Legion::LLM::RateLimitError => e
    handle_exception(e, level: :warn, handled: true, operation: 'llm.api.namespaces.openai.completions.rate_limit')
    openai_error(e.message, type: 'rate_limit_error', code: 'rate_limit_exceeded', status_code: 429)
  rescue Legion::LLM::ProviderDown, Legion::LLM::ProviderError => e
    handle_exception(e, level: :error, handled: true, operation: 'llm.api.namespaces.openai.completions.provider')
    openai_error(e.message, type: 'server_error', status_code: 502)
  rescue StandardError => e
    handle_exception(e, level: :error, handled: false, operation: 'llm.api.namespaces.openai.completions')
    openai_error(e.message, type: 'server_error', status_code: 500)
  end
  # rubocop:enable Metrics/BlockLength

  log.debug('[llm][api][namespaces][openai][completions] routes registered')
rescue StandardError => e
  handle_exception(e, level: :error, handled: false, operation: 'llm.api.namespaces.openai.completions.register')
end