Module: Legion::LLM::API::Namespaces::Native::Chat

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

Constant Summary collapse

ASYNC_POOL = 
Concurrent::FixedThreadPool.new(
  [4, (Concurrent.processor_count / 2)].max,
  fallback_policy: :abort
)

Class Method Summary collapse

Class Method Details

.registered(ns_context) ⇒ Object

rubocop:disable Metrics/AbcSize,Metrics/MethodLength



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
 
# File 'lib/legion/llm/api/namespaces/native/chat.rb', line 29

def self.registered(ns_context) # rubocop:disable Metrics/AbcSize,Metrics/MethodLength
  log.debug('[llm][api][namespaces][chat] registering routes')

  ns_context.post '' do # rubocop:disable Metrics/BlockLength
    log.debug("[llm][api][namespaces][chat] action=received params=#{params.keys}")
    require_llm!

    body = parse_request_body
    validate_required!(body, :message)

    message    = body[:message]
    request_id = body[:request_id] || SecureRandom.uuid
    model      = body[:model]
    provider   = body[:provider]

    log.debug("[llm][api][namespaces][chat] action=dispatch request_id=#{request_id} model=#{model || 'auto'}")

    cache_active = defined?(Legion::LLM::Cache::Response) &&
                   Legion::LLM::Cache::Response.respond_to?(:init_request) &&
                   env['HTTP_X_LEGION_SYNC'] != 'true'

    if cache_active
      log.debug("[llm][api][namespaces][chat] action=async_dispatch request_id=#{request_id}")
      Legion::LLM::Cache::Response.init_request(request_id)
      llm = Legion::LLM
      rc  = Legion::LLM::Cache::Response

      begin
        ASYNC_POOL.post do
          session  = llm.chat_direct(model: model, provider: provider)
          response = session.ask(message)
          rc.complete(
            request_id,
            response: response.content,
            meta:     {
              model:      session.model.to_s,
              tokens_in:  response.respond_to?(:input_tokens) ? response.input_tokens : nil,
              tokens_out: response.respond_to?(:output_tokens) ? response.output_tokens : nil
            }
          )
          log.debug("[llm][api][namespaces][chat] action=async_complete request_id=#{request_id}")
        rescue StandardError => e
          handle_exception(e, level: :error, handled: true, operation: 'llm.api.chat.async', request_id: request_id)
          rc.fail_request(request_id, code: 'llm_error', message: e.message)
        end
      rescue Concurrent::RejectedExecutionError
        log.warn("[llm][api][namespaces][chat] action=async_pool_saturated request_id=#{request_id}")
        halt json_error('queue_full', 'Chat queue is currently full. Please retry.', status_code: 503)
      end

      log.info("[llm][api][namespaces][chat] action=queued request_id=#{request_id}")
      json_response({ request_id: request_id, poll_key: "llm:#{request_id}:status" }, status_code: 202)
    else
      log.debug("[llm][api][namespaces][chat] action=sync_dispatch request_id=#{request_id}")
      result = Legion::LLM.chat(
        message:  message,
        model:    model,
        provider: provider,
        caller:   build_server_caller(source: 'api', path: request.path, env: env)
      )

      if result.is_a?(Legion::LLM::Inference::Response)
        raw_msg         = result.message
        content         = raw_msg.is_a?(Hash) ? (raw_msg[:content] || raw_msg['content']) : raw_msg.to_s
        routing         = result.routing || {}
        resolved_model  = routing[:model] || routing['model']
        tokens          = result.tokens || {}
        log.info("[llm][api][namespaces][chat] action=completed request_id=#{request_id} model=#{resolved_model}")
        json_response(
          {
            response: content,
            meta:     {
              model:      resolved_model.to_s,
              tokens_in:  token_value(tokens, :input),
              tokens_out: token_value(tokens, :output)
            }
          },
          status_code: 201
        )
      else
        response = result
        log.info("[llm][api][namespaces][chat] action=completed request_id=#{request_id} result_class=#{response.class}")
        json_response(
          {
            response: response.respond_to?(:content) ? response.content : response.to_s,
            meta:     {
              model:      response.respond_to?(:model_id) ? response.model_id.to_s : model.to_s,
              tokens_in:  response.respond_to?(:input_tokens) ? response.input_tokens : nil,
              tokens_out: response.respond_to?(:output_tokens) ? response.output_tokens : nil
            }
          },
          status_code: 201
        )
      end
    end
  end

  log.debug('[llm][api][namespaces][chat] routes registered')
end