Class: GitFit::LLM::Client

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

Overview

OpenAI-compatible chat-completions client (POST #endpoint/chat/completions).

Provider-agnostic: endpoint / model / api key all come from the config ai: section — the gem never hardcodes a vendor. Non-streaming (P1); SSE can be layered later without changing call sites.

Instance Attribute Summary collapse

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(endpoint:, model:, api_key:, timeout: 60) ⇒ Client

Returns a new instance of Client.



43
44
45
46
47
48
 
# File 'lib/git_fit/llm/client.rb', line 43

def initialize(endpoint:, model:, api_key:, timeout: 60)
  @endpoint = endpoint.to_s.sub(%r{/+\z}, '')
  @model = model
  @api_key = api_key
  @timeout = timeout.is_a?(Numeric) ? timeout : 60
end

Instance Attribute Details

#endpoint ⇒ Object (readonly)

Returns the value of attribute endpoint.



14
15
16
 
# File 'lib/git_fit/llm/client.rb', line 14

def endpoint
  @endpoint
end

#model ⇒ Object (readonly)

Returns the value of attribute model.



14
15
16
 
# File 'lib/git_fit/llm/client.rb', line 14

def model
  @model
end

#timeout ⇒ Object (readonly)

Returns the value of attribute timeout.



14
15
16
 
# File 'lib/git_fit/llm/client.rb', line 14

def timeout
  @timeout
end

Class Method Details

.from_config(config) ⇒ Object

Builds a client from GitFit::Config. Returns nil when ai.enabled is false (caller decides to warn+skip); raises ConfigError with a Fix: hint when enabled but misconfigured.



19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
 
# File 'lib/git_fit/llm/client.rb', line 19

def self.from_config(config)
  ai = config.ai_config
  return nil unless ai['enabled']

  missing = %w[endpoint model].select { |k| ai[k].to_s.strip.empty? }
  unless missing.empty?
    raise ConfigError,
          "LLM: ai.#{missing.join(', ai.')} not configured\n" \
          '  Fix: set ai.endpoint / ai.model in config.yml ' \
          '(e.g. https://open.bigmodel.cn/api/paas/v4 + glm-4-flash)'
  end

  key_env = ai['api_key_env'].to_s.strip
  key_env = 'AI_API_KEY' if key_env.empty?
  api_key = ENV[key_env].to_s
  if api_key.empty?
    raise ConfigError,
          "LLM: env #{key_env} is empty\n" \
          "  Fix: export #{key_env}=<your-api-key> (local: add to .env.local and source it)"
  end

  new(endpoint: ai['endpoint'], model: ai['model'], api_key: api_key, timeout: ai['timeout'])
end

Instance Method Details

#chat(messages:, temperature: nil) ⇒ Object

messages: [{ 'role' => 'system'|'user', 'content' => '...' }, ...] Returns the assistant content string.



52
53
54
55
56
57
 
# File 'lib/git_fit/llm/client.rb', line 52

def chat(messages:, temperature: nil)
  payload = { model: @model, messages: messages }
  payload[:temperature] = temperature if temperature
  status, body = http_post(chat_completions_url, JSON.generate(payload))
  parse_content(status, body)
end