Class: CompletionKit::OllamaClient

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

Constant Summary

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 



43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
 
# File 'app/services/completion_kit/ollama_client.rb', line 43

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

  response = build_connection(api_endpoint).get("/v1/models") do |req|
    req.headers["Authorization"] = "Bearer #{api_key}" if api_key.present?
  end

  return [] unless response.success?

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

#configuration_errors ⇒ Object 



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

def configuration_errors
  errors = []
  errors << "Ollama API endpoint is not configured" unless api_endpoint.present?
  errors
end

#configured? ⇒ Boolean

Returns:

  • (Boolean) 


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

def configured?
  api_endpoint.present?
end

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



4
5
6
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
 
# File 'app/services/completion_kit/ollama_client.rb', line 4

def generate_completion(prompt, options = {})
  @temperature_dropped = false
  return "Error: API endpoint not 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 = resolve_temperature(options)

  response = post_completion(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_completion(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: "ollama",
      status: 429,
      retry_after: nil
    )
  end

  if response.success?
    data = JSON.parse(response.body)
    data["choices"][0]["text"].strip
  else
    "Error: #{response.status} - #{response.body}"
  end
rescue CompletionKit::RateLimitError
  raise
rescue Faraday::Error
  raise
rescue => e
  "Error: #{e.message}"
end