Module: RubyLLM::Protocols::Cohere::Transcription

Defined in:
lib/ruby_llm/protocols/cohere/transcription.rb

Overview

Audio transcription methods for the Cohere v2 API integration

Constant Summary collapse

DEFAULT_LANGUAGE = 
'en'

Class Method Summary collapse

Class Method Details

.parse_transcription_response(response, model:) ⇒ Object

Cohere Transcribe returns the transcript alone: no timestamps, segments, or speaker labels.



32
33
34
35
36
37
 
# File 'lib/ruby_llm/protocols/cohere/transcription.rb', line 32

def parse_transcription_response(response, model:)
  data = response.body
  return RubyLLM::Transcription.new(text: data, model: model) if data.is_a?(String)

  RubyLLM::Transcription.new(text: data['text'], model: model)
end

.render_transcription_payload(file_part, model:, language:, format: nil, speaker_names: nil, speaker_references: nil, provider_options: {}, prompt: nil, temperature: nil) ⇒ Object

Cohere parses the multipart body as it streams, so every form field has to be written before the file part or the request is rejected as missing them. The file goes in last, after provider options. rubocop:disable-next Lint/UnusedMethodArgument



20
21
22
23
24
25
26
27
28
 
# File 'lib/ruby_llm/protocols/cohere/transcription.rb', line 20

def render_transcription_payload(file_part, model:, language:, format: nil, speaker_names: nil,
                                 speaker_references: nil, provider_options: {}, prompt: nil,
                                 temperature: nil)
  {
    model: model,
    language: language || DEFAULT_LANGUAGE,
    temperature: temperature
  }.compact.merge(provider_options).merge(file: file_part)
end

.transcription_url ⇒ Object 



12
13
14
 
# File 'lib/ruby_llm/protocols/cohere/transcription.rb', line 12

def transcription_url
  'v2/audio/transcriptions'
end