Module: RubyLLM::Providers::OpenAI::Chat

Included in:
RubyLLM::Providers::OpenAI
Defined in:
lib/ruby_llm/providers/openai/chat.rb

Overview

Chat methods of the OpenAI API integration

Class Method Summary collapse

Instance Method Summary collapse

Class Method Details

.cache_read_tokens(usage) ⇒ Object 



112
113
114
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 112

def cache_read_tokens(usage)
  usage.dig('prompt_tokens_details', 'cached_tokens') || usage['prompt_cache_hit_tokens']
end

.cache_write_tokens(usage) ⇒ Object 



116
117
118
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 116

def cache_write_tokens(usage)
  usage.dig('prompt_tokens_details', 'cache_write_tokens') || 0
end

.extract_content_and_thinking(content) ⇒ Object 



175
176
177
178
179
180
181
182
183
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 175

def extract_content_and_thinking(content)
  return extract_think_tag_content(content) if content.is_a?(String)
  return [content, nil] unless content.is_a?(Array)

  text = extract_text_from_blocks(content)
  thinking = extract_thinking_from_blocks(content)

  [text.empty? ? nil : text, thinking.empty? ? nil : thinking]
end

.extract_text_from_blocks(blocks) ⇒ Object 



185
186
187
188
189
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 185

def extract_text_from_blocks(blocks)
  blocks.filter_map do |block|
    block['text'] if block['type'] == 'text' && block['text'].is_a?(String)
  end.join
end

.extract_think_tag_content(text) ⇒ Object 



210
211
212
213
214
215
216
217
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 210

def extract_think_tag_content(text)
  return [text, nil] unless text.include?('<think>')

  thinking = text.scan(%r{<think>(.*?)</think>}m).join
  content = text.gsub(%r{<think>.*?</think>}m, '').strip

  [content.empty? ? nil : content, thinking.empty? ? nil : thinking]
end

.extract_thinking_from_blocks(blocks) ⇒ Object 



191
192
193
194
195
196
197
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 191

def extract_thinking_from_blocks(blocks)
  blocks.filter_map do |block|
    next unless block['type'] == 'thinking'

    extract_thinking_text_from_block(block)
  end.join
end

.extract_thinking_signature(message_data) ⇒ Object 



170
171
172
173
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 170

def extract_thinking_signature(message_data)
  candidate = message_data['reasoning_signature'] || message_data['signature']
  candidate.is_a?(String) ? candidate : nil
end

.extract_thinking_text(message_data) ⇒ Object 



165
166
167
168
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 165

def extract_thinking_text(message_data)
  candidate = message_data['reasoning_content'] || message_data['reasoning'] || message_data['thinking']
  candidate.is_a?(String) ? candidate : nil
end

.extract_thinking_text_from_block(block) ⇒ Object 



199
200
201
202
203
204
205
206
207
208
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 199

def extract_thinking_text_from_block(block)
  thinking_block = block['thinking']
  return thinking_block if thinking_block.is_a?(String)

  if thinking_block.is_a?(Array)
    return thinking_block.filter_map { |item| item['text'] if item['type'] == 'text' }.join
  end

  block['text'] if block['text'].is_a?(String)
end

.format_messages(messages) ⇒ Object 



124
125
126
127
128
129
130
131
132
133
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 124

def format_messages(messages)
  messages.map do |msg|
    {
      role: format_role(msg.role),
      content: Media.format_content(msg.content),
      tool_calls: format_tool_calls(msg.tool_calls),
      tool_call_id: msg.tool_call_id
    }.compact.merge(format_thinking(msg))
  end
end

.format_role(role) ⇒ Object 



135
136
137
138
139
140
141
142
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 135

def format_role(role)
  case role
  when :system
    @config.openai_use_system_role ? 'system' : 'developer'
  else
    role.to_s
  end
end

.format_thinking(msg) ⇒ Object 



150
151
152
153
154
155
156
157
158
159
160
161
162
163
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 150

def format_thinking(msg)
  return {} unless msg.role == :assistant

  thinking = msg.thinking
  return {} unless thinking

  payload = {}
  if thinking.text
    payload[:reasoning] = thinking.text
    payload[:reasoning_content] = thinking.text
  end
  payload[:reasoning_signature] = thinking.signature if thinking.signature
  payload
end

.generated_tokens_from_total(usage) ⇒ Object 



104
105
106
107
108
109
110
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 104

def generated_tokens_from_total(usage)
  prompt_tokens = usage['prompt_tokens']
  total_tokens = usage['total_tokens']
  return unless prompt_tokens && total_tokens

  [total_tokens.to_i - prompt_tokens.to_i, 0].max
end

.input_tokens(usage) ⇒ Object 



84
85
86
87
88
89
90
91
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 84

def input_tokens(usage)
  return usage['prompt_cache_miss_tokens'] if usage['prompt_cache_miss_tokens']

  prompt_tokens = usage['prompt_tokens']
  return unless prompt_tokens

  [prompt_tokens.to_i - cache_read_tokens(usage).to_i - cache_write_tokens(usage).to_i, 0].max
end

.output_tokens(usage) ⇒ Object 



93
94
95
96
97
98
99
100
101
102
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 93

def output_tokens(usage)
  completion_tokens = usage['completion_tokens']
  return unless completion_tokens

  completion_tokens = completion_tokens.to_i
  generated_tokens = generated_tokens_from_total(usage)
  return completion_tokens unless generated_tokens && generated_tokens > completion_tokens

  generated_tokens
end

.parse_completion_response(response) ⇒ Object

rubocop:enable Metrics/ParameterLists,Metrics/PerceivedComplexity

Raises:



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
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 54

def parse_completion_response(response)
  data = response.body
  return if data.empty?

  raise Error.new(response, data.dig('error', 'message')) if data.dig('error', 'message')

  message_data = data.dig('choices', 0, 'message')
  return unless message_data

  usage = data['usage'] || {}
  thinking_tokens = thinking_tokens(usage)
  content, thinking_from_blocks = extract_content_and_thinking(message_data['content'])
  thinking_text = thinking_from_blocks || extract_thinking_text(message_data)
  thinking_signature = extract_thinking_signature(message_data)

  Message.new(
    role: :assistant,
    content: content,
    thinking: Thinking.build(text: thinking_text, signature: thinking_signature),
    tool_calls: parse_tool_calls(message_data['tool_calls']),
    input_tokens: input_tokens(usage),
    output_tokens: output_tokens(usage),
    cached_tokens: cache_read_tokens(usage),
    cache_creation_tokens: cache_write_tokens(usage),
    thinking_tokens: thinking_tokens,
    model_id: data['model'],
    raw: response
  )
end

.render_payload(messages, tools:, temperature:, model:, stream: false, schema: nil, thinking: nil, tool_prefs: nil) ⇒ Object

rubocop:disable Metrics/ParameterLists,Metrics/PerceivedComplexity



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
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 15

def render_payload(messages, tools:, temperature:, model:, stream: false, schema: nil,
                   thinking: nil, tool_prefs: nil)
  tool_prefs ||= {}
  payload = {
    model: model.id,
    messages: format_messages(messages),
    stream: stream
  }

  payload[:temperature] = temperature unless temperature.nil?
  if tools.any?
    payload[:tools] = tools.map { |_, tool| tool_for(tool) }
    payload[:tool_choice] = build_tool_choice(tool_prefs[:choice]) unless tool_prefs[:choice].nil?
    payload[:parallel_tool_calls] = tool_prefs[:calls] == :many unless tool_prefs[:calls].nil?
  end

  if schema
    schema_name = schema[:name]
    schema_def = schema[:schema]
    strict = schema[:strict]

    payload[:response_format] = {
      type: 'json_schema',
      json_schema: {
        name: schema_name,
        schema: schema_def,
        strict: strict
      }
    }
  end

  effort = resolve_effort(thinking)
  payload[:reasoning_effort] = effort if effort

  payload[:stream_options] = { include_usage: true } if stream
  payload
end

.resolve_effort(thinking) ⇒ Object 



144
145
146
147
148
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 144

def resolve_effort(thinking)
  return nil unless thinking

  thinking.respond_to?(:effort) ? thinking.effort : thinking
end

.thinking_tokens(usage) ⇒ Object 



120
121
122
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 120

def thinking_tokens(usage)
  usage.dig('completion_tokens_details', 'reasoning_tokens') || usage['reasoning_tokens']
end

Instance Method Details

#completion_url ⇒ Object 



8
9
10
 
# File 'lib/ruby_llm/providers/openai/chat.rb', line 8

def completion_url
  'chat/completions'
end