Class: Prescient::Provider::OpenAI

Inherits:
Base
  • Object
show all
Includes:
HTTParty
Defined in:
lib/prescient/provider/openai.rb

Overview

OpenAI API provider adapter.

Constant Summary collapse

EMBEDDING_DIMENSIONS =

Known embedding dimensions for OpenAI embedding models.

{
  'text-embedding-3-small' => 1536,
  'text-embedding-3-large' => 3072,
  'text-embedding-ada-002' => 1536,
}.freeze

Instance Attribute Summary

Attributes inherited from Base

#options, #provider_name

Instance Method Summary collapse

Methods inherited from Base

#available?, #build_prompt, #clean_text, #default_context_configs, #default_prompt_templates, #extract_embedding_text, #extract_text_values, #format_context_item, #handle_errors, #validate_embedding_dimensions

Constructor Details

#initialize(**options) ⇒ OpenAI

Returns a new instance of OpenAI.



18
19
20
21
22
 
# File 'lib/prescient/provider/openai.rb', line 18

def initialize(**options)
  super
  @provider_name = 'OpenAI'
  self.class.default_timeout(@options[:timeout] || 60)
end

Instance Method Details

#generate_embedding(text, **options) ⇒ Array<Float>

Generate an embedding through the OpenAI embeddings API.

Parameters:

  • text (String) —

    Text to embed

Returns:

  • (Array<Float>) —

    Embedding vector



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
54
55
56
 
# File 'lib/prescient/provider/openai.rb', line 27

def generate_embedding(text, **options)
  handle_errors do
    clean_text_input = clean_text(text)

    embedding_model = options[:model] || @options[:embedding_model]
    response = self.class.post('/v1/embeddings',
                               headers: {
                                 'Content-Type'  => 'application/json',
                                 'Authorization' => "Bearer #{@options[:api_key]}",
                               },
                               body:    {
                                 model:           embedding_model,
                                 input:           clean_text_input,
                                 encoding_format: 'float',
                               }.to_json)

    validate_response!(response, 'embedding generation')

    embedding_data = response.parsed_response.dig('data', 0, 'embedding')
    raise Prescient::InvalidResponseError, 'No embedding returned' unless embedding_data

    expected_dimensions = EMBEDDING_DIMENSIONS[embedding_model] || @options[:embedding_dimensions]
    unless expected_dimensions
      raise Prescient::Error,
            "Embedding dimensions are required for model #{embedding_model}"
    end

    validate_embedding_dimensions(embedding_data, expected_dimensions)
  end
end

#generate_response(prompt, context_items = [], **options) ⇒ Hash

Generate a response through the OpenAI chat completions API.

Parameters:

  • prompt (String) —

    Prompt to send

  • context_items (Array<Hash, String>) (defaults to: []) —

    Optional context items

Returns:

  • (Hash) —

    Normalized response data



62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
 
# File 'lib/prescient/provider/openai.rb', line 62

def generate_response(prompt, context_items = [], **options)
  handle_errors do
    formatted_prompt = build_prompt(prompt, context_items)

    response = self.class.post('/v1/chat/completions',
                               headers: {
                                 'Content-Type'  => 'application/json',
                                 'Authorization' => "Bearer #{@options[:api_key]}",
                               },
                               body:    {
                                 model:       options[:model] || @options[:chat_model],
                                 messages:    [
                                   {
                                     role:    'user',
                                     content: formatted_prompt,
                                   },
                                 ],
                                 max_tokens:  options[:max_tokens] || 2000,
                                 temperature: options[:temperature] || 0.7,
                                 top_p:       options[:top_p] || 0.9,
                               }.to_json)

    validate_response!(response, 'text generation')

    puts "response.parsed_response: #{response.parsed_response.inspect}"

    content = response.parsed_response.dig('choices', 0, 'message', 'content')
    raise Prescient::InvalidResponseError, 'No response generated' unless content

    {
      response:        content.strip,
      model:           options[:model] || @options[:chat_model],
      provider:        'openai',
      processing_time: nil,
      metadata:        {
        usage:         response.parsed_response['usage'],
        finish_reason: response.parsed_response.dig('choices', 0, 'finish_reason'),
      },
    }
  end
end

#health_check ⇒ Hash

Check OpenAI model availability via /v1/models.

reachable indicates the API answered successfully. ready indicates that both configured models appear in the returned model list.

Returns:

  • (Hash) —

    Provider health information



110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
 
# File 'lib/prescient/provider/openai.rb', line 110

def health_check
  handle_errors do
    response = self.class.get('/v1/models',
                              headers: {
                                'Authorization' => "Bearer #{@options[:api_key]}",
                              })

    if response.success?
      models = response.parsed_response['data'] || []
      embedding_available = models.any? { |m| m['id'] == @options[:embedding_model] }
      chat_available = models.any? { |m| m['id'] == @options[:chat_model] }

      {
        status:           'healthy',
        provider:         'openai',
        reachable:        true,
        models_available: models.map { |m| m['id'] },
        embedding_model:  {
          name:      @options[:embedding_model],
          available: embedding_available,
        },
        chat_model:       {
          name:      @options[:chat_model],
          available: chat_available,
        },
        ready:            embedding_available && chat_available,
      }
    else
      {
        status:    'unhealthy',
        provider:  'openai',
        reachable: true,
        error:     "HTTP #{response.code}",
        message:   response.message,
        ready:     false,
      }
    end
  end
rescue Prescient::Error => e
  {
    status:    'unavailable',
    provider:  'openai',
    reachable: false,
    error:     e.class.name,
    message:   e.message,
    ready:     false,
  }
end

#list_models ⇒ Array<Hash>

List models available to the configured OpenAI account.

Returns:

  • (Array<Hash>) —

    Model descriptors



162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
 
# File 'lib/prescient/provider/openai.rb', line 162

def list_models
  handle_errors do
    response = self.class.get('/v1/models',
                              headers: {
                                'Authorization' => "Bearer #{@options[:api_key]}",
                              })
    validate_response!(response, 'model listing')

    models = response.parsed_response['data'] || []
    models.map do |model|
      {
        name:     model['id'],
        created:  model['created'],
        owned_by: model['owned_by'],
      }
    end
  end
end

#validate_configuration! ⇒ Object (protected)

Raises:



183
184
185
186
187
188
189
190
 
# File 'lib/prescient/provider/openai.rb', line 183

def validate_configuration!
  required_options = [:api_key, :embedding_model, :chat_model]
  missing_options = required_options.select { |opt| @options[opt].nil? }

  return unless missing_options.any?

  raise Prescient::Error, "Missing required options: #{missing_options.join(', ')}"
end