Class: CompletionKit::OpenRouterClient

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

Constant Summary collapse

BASE_URL = 
"https://openrouter.ai".freeze
REFERER = 
"https://completionkit.com".freeze
APP_TITLE = 
"CompletionKit".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 



52
53
54
 
# File 'app/services/completion_kit/open_router_client.rb', line 52

def available_models
  []
end

#configuration_errors ⇒ Object 



60
61
62
63
64
 
# File 'app/services/completion_kit/open_router_client.rb', line 60

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

#configured? ⇒ Boolean

Returns:

  • (Boolean) 


56
57
58
 
# File 'app/services/completion_kit/open_router_client.rb', line 56

def configured?
  api_key.present?
end

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



8
9
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
 
# File 'app/services/completion_kit/open_router_client.rb', line 8

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

  model = options[:model] || "openai/gpt-4o-mini"
  max_tokens = options[:max_tokens] || 8192
  temperature = resolve_temperature(options)

  response = post_chat(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_chat(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: "openrouter",
      status: 429,
      retry_after: response.headers && response.headers["Retry-After"]&.to_i
    )
  end

  if response.success?
    data = JSON.parse(response.body)
    choice = data.dig("choices", 0) || {}
    if choice["finish_reason"] == "length"
      return "Error: response truncated by max_tokens=#{max_tokens} before visible content was emitted (reasoning model burned through the budget)"
    end
    content = choice.dig("message", "content").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
  raise
rescue => e
  "Error: #{e.message}"
end