Class: CompletionKit::AzureFoundryClient

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

Instance Method Summary collapse

Methods inherited from LlmClient

for_model, for_provider, #initialize

Constructor Details

This class inherits a constructor from CompletionKit::LlmClient

Instance Method Details

#available_models ⇒ Object 



48
49
50
51
52
53
54
55
56
57
58
59
60
 
# File 'app/services/completion_kit/azure_foundry_client.rb', line 48

def available_models
  return [] unless configured?
  return [] unless ProviderEndpoint.safe?(api_endpoint)

  response = build_connection(azure_base_url).get("/openai/deployments?api-version=#{api_version}") do |req|
    req.headers["api-key"] = api_key
  end
  return [] unless response.success?

  JSON.parse(response.body).fetch("data", []).map { |entry| { id: entry["id"], name: entry["id"] } }
rescue StandardError
  []
end

#configuration_errors ⇒ Object 



66
67
68
69
70
71
72
 
# File 'app/services/completion_kit/azure_foundry_client.rb', line 66

def configuration_errors
  errors = []
  errors << "Azure endpoint is not configured" if api_endpoint.blank?
  errors << "Azure API key is not configured" if api_key.blank?
  errors << "Azure api-version is not configured" if api_version.blank?
  errors
end

#configured? ⇒ Boolean

Returns:

  • (Boolean) 


62
63
64
 
# File 'app/services/completion_kit/azure_foundry_client.rb', line 62

def configured?
  configuration_errors.empty?
end

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



7
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
 
# File 'app/services/completion_kit/azure_foundry_client.rb', line 7

def generate_completion(prompt, options = {})
  @temperature_dropped = false
  return "Error: Azure provider is not fully configured" unless configured?
  return "Error: API endpoint resolves to a private address" unless ProviderEndpoint.safe?(api_endpoint)

  model = options[:model]
  max_tokens = options[:max_tokens] || 1000
  temperature = options[:temperature] || 0.7

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

  if response.status == 400 && 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: "azure_foundry",
      status: 429,
      retry_after: response.headers["Retry-After"]&.to_i
    )
  end

  if response.success?
    data = JSON.parse(response.body)
    content = data.dig("choices", 0, "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

#temperature_dropped? ⇒ Boolean

Returns:

  • (Boolean) 


3
4
5
 
# File 'app/services/completion_kit/azure_foundry_client.rb', line 3

def temperature_dropped?
  @temperature_dropped == true
end