Module: RubyLLM::Providers::Vllm::Chat

Included in:
RubyLLM::Providers::Vllm
Defined in:
lib/legion/llm/patches/ruby_llm_vllm.rb

Instance Method Summary collapse

Instance Method Details

#format_messages(messages) ⇒ Object 



11
12
13
14
15
16
17
18
19
20
 
# File 'lib/legion/llm/patches/ruby_llm_vllm.rb', line 11

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

#format_role(role) ⇒ Object 



7
8
9
 
# File 'lib/legion/llm/patches/ruby_llm_vllm.rb', line 7

def format_role(role)
  role.to_s
end

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



22
23
24
25
26
27
28
29
30
31
32
 
# File 'lib/legion/llm/patches/ruby_llm_vllm.rb', line 22

def render_payload(messages, tools:, temperature:, model:, stream: false, schema: nil,
                   thinking: nil, tool_prefs: nil)
  payload = super
  enable = if thinking.nil?
             vllm_thinking_default
           else
             thinking ? true : false
           end
  payload[:chat_template_kwargs] = { enable_thinking: enable }
  payload
end