Class: CompletionKit::OpenAiClient

Inherits:
LlmClient
  • Object
show all
Defined in:
app/services/completion_kit/open_ai_client.rb

Constant Summary collapse

STATIC_MODELS = 
[
  { id: "gpt-5.4-mini", name: "GPT-5.4 Mini" },
  { id: "gpt-4.1-mini", name: "GPT-4.1 Mini" },
  { id: "gpt-4o-mini", name: "GPT-4o Mini" }
].freeze

Constants inherited from LlmClient

LlmClient::DEFAULT_TEMPERATURE, LlmClient::REFUSAL_PHRASE, LlmClient::TEMPERATURE_REFUSAL

Instance Method Summary collapse

Methods inherited from LlmClient

for_model, for_provider, #initialize, #temperature_dropped?

Constructor Details

This class inherits a constructor from CompletionKit::LlmClient

Instance Method Details

#available_models ⇒ Object 



55
56
57
 
# File 'app/services/completion_kit/open_ai_client.rb', line 55

def available_models
  STATIC_MODELS
end

#configuration_errors ⇒ Object 



63
64
65
66
67
 
# File 'app/services/completion_kit/open_ai_client.rb', line 63

def configuration_errors
  errors = []
  errors << "OpenAI API key is not configured" unless api_key.present?
  errors
end

#configured? ⇒ Boolean

Returns:

  • (Boolean) 


59
60
61
 
# File 'app/services/completion_kit/open_ai_client.rb', line 59

def configured?
  api_key.present?
end

#generate_completion(prompt, options = {}) ⇒ Object 



10
11
12
13
14
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
52
53
 
# File 'app/services/completion_kit/open_ai_client.rb', line 10

def generate_completion(prompt, options = {})
  @temperature_dropped = false
  return "Error: API key not configured" unless configured?

  model = options[:model] || "gpt-4.1-mini"
  max_tokens = options[:max_tokens] || 8192
  temperature = resolve_temperature(options)

  response = post_responses(model: model, prompt: prompt, max_tokens: max_tokens, temperature: temperature)

  if response.status == 400 && !temperature.nil? && temperature_unsupported?(response.body)
    @temperature_dropped = true
    response = post_responses(model: model, prompt: prompt, max_tokens: max_tokens, temperature: nil)
  end

  if response.status == 429
    raise CompletionKit::RateLimitError.new(
      response.body.to_s.truncate(500),
      provider: "openai",
      status: 429,
      retry_after: response.headers && response.headers["Retry-After"]&.to_i
    )
  end

  if response.success?
    data = JSON.parse(response.body)
    if data["status"] == "incomplete"
      reason = data.dig("incomplete_details", "reason") || "unknown"
      return "Error: response incomplete (#{reason}). Increase max_tokens=#{max_tokens} or pick a non-reasoning judge model"
    end
    message = Array(data["output"]).find { |o| o["type"] == "message" }
    content = message&.dig("content", 0, "text").to_s.strip
    return "Error: model returned empty content" if content.empty?
    content
  else
    "Error: #{response.status} - #{response.body}"
  end
rescue CompletionKit::RateLimitError
  raise
rescue Faraday::Error => e
  raise
rescue => e
  "Error: #{e.message}"
end