Class: Slidict::Llm::Client

Inherits:
Object
  • Object
show all
Defined in:
lib/slidict/llm/client.rb

Overview

Talks to any OpenAI Compatible API (OpenAI, Ollama, LM Studio, vLLM, etc.) via the standard /chat/completions endpoint. Configure the target with Slidict::Config (base_url, api_key, model).

Defined Under Namespace

Classes: Error

Instance Method Summary collapse

Constructor Details

#initialize(base_url:, api_key:, model:) ⇒ Client

Returns a new instance of Client.



15
16
17
18
19
 
# File 'lib/slidict/llm/client.rb', line 15

def initialize(base_url:, api_key:, model:)
  @base_url = base_url
  @api_key = api_key
  @model = model
end

Instance Method Details

#generate_slides(deck) ⇒ Object 



32
33
34
35
 
# File 'lib/slidict/llm/client.rb', line 32

def generate_slides(deck)
  content = chat_completion(prompt_for(deck))
  slides_from(content)
end

#lint_slides(slide_texts) ⇒ Object

slide_texts is an array of slide bodies (1-indexed by position) as produced by Slidict::Lint::SlideParser. Returns an array of Slidict::Lint::Finding.



40
41
42
43
 
# File 'lib/slidict/llm/client.rb', line 40

def lint_slides(slide_texts)
  content = chat_completion(lint_prompt_for(slide_texts))
  findings_from(content)
end

#verify_connection! ⇒ Object

Checks that the endpoint is reachable before the (slower, more expensive) chat completion request is made. Raises Error on failure.

Raises:



23
24
25
26
27
28
29
30
 
# File 'lib/slidict/llm/client.rb', line 23

def verify_connection!
  uri = endpoint_uri("models")
  request = Net::HTTP::Get.new(uri)
  request["Authorization"] = "Bearer #{@api_key}"
  response = perform_request(uri, request)

  raise Error, "#{response.code} #{response.message}" unless response.is_a?(Net::HTTPSuccess)
end