Class: Mistral::Client

Inherits:
ClientBase show all
Defined in:
lib/mistral/client.rb

Overview

Synchronous wrapper around the async client

Instance Attribute Summary

Attributes inherited from ClientBase

#api_key, #endpoint, #max_retries, #timeout

Instance Method Summary collapse

Constructor Details

#initialize(api_key: nil, endpoint: ENDPOINT, max_retries: 5, timeout: 120) ⇒ Client

Returns a new instance of Client.



8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
 
# File 'lib/mistral/client.rb', line 8

def initialize(
  api_key: nil,
  endpoint: ENDPOINT,
  max_retries: 5,
  timeout: 120
)
  super(endpoint: endpoint, api_key: api_key, max_retries: max_retries, timeout: timeout)

  @client = HTTP.persistent(ENDPOINT)
                .follow
                .timeout(timeout)
                .use(:line_iterable_body)
                .headers('Accept' => 'application/json',
                  'User-Agent' => "mistral-client-ruby/#{VERSION}",
                  'Authorization' => "Bearer #{@api_key}",
                  'Content-Type' => 'application/json'
                )
end

Instance Method Details

#chat(messages:, model: nil, tools: nil, temperature: nil, max_tokens: nil, top_p: nil, random_seed: nil, safe_mode: false, safe_prompt: false, tool_choice: nil, response_format: nil) ⇒ ChatCompletionResponse

A chat endpoint that returns a single response.

Parameters:

  • messages (Array<ChatMessage>) —

    An array of messages to chat with, e.g.

    ‘user’, content: ‘What is the best French cheese?’
  • model (String) (defaults to: nil) —

    The name of the model to chat with, e.g. mistral-tiny

  • tools (Array<Hash>) (defaults to: nil) —

    A list of tools to use.

  • temperature (Float) (defaults to: nil) —

    The temperature to use for sampling, e.g. 0.5.

  • max_tokens (Integer) (defaults to: nil) —

    The maximum number of tokens to generate, e.g. 100.

  • top_p (Float) (defaults to: nil) —

    The cumulative probability of tokens to generate, e.g. 0.9.

  • random_seed (Integer) (defaults to: nil) —

    The random seed to use for sampling, e.g. 42.

  • safe_mode (Boolean) (defaults to: false) —

    Deprecated, use safe_prompt instead.

  • safe_prompt (Boolean) (defaults to: false) —

    Whether to use safe prompt, e.g. true.

  • tool_choice (String, ToolChoice) (defaults to: nil) —

    The tool choice.

  • response_format (Hash<String, String>, ResponseFormat) (defaults to: nil) —

    The response format.

Returns:

Raises:



43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
 
# File 'lib/mistral/client.rb', line 43

def chat(
  messages:,
  model: nil,
  tools: nil,
  temperature: nil,
  max_tokens: nil,
  top_p: nil,
  random_seed: nil,
  safe_mode: false,
  safe_prompt: false,
  tool_choice: nil,
  response_format: nil
)
  request = make_chat_request(
    messages: messages,
    model: model,
    tools: tools,
    temperature: temperature,
    max_tokens: max_tokens,
    top_p: top_p,
    random_seed: random_seed,
    stream: false,
    safe_prompt: safe_mode || safe_prompt,
    tool_choice: tool_choice,
    response_format: response_format
  )

  single_response = request('post', 'v1/chat/completions', json: request)

  single_response.each do |response|
    return ChatCompletionResponse.new(response)
  end

  raise Mistral::Error.new(message: 'No response received')
end

#chat_stream(messages:, model: nil, tools: nil, temperature: nil, max_tokens: nil, top_p: nil, random_seed: nil, safe_mode: false, safe_prompt: false, tool_choice: nil, response_format: nil) ⇒ Enumerator<ChatCompletionStreamResponse>

A chat endpoint that streams responses.

Parameters:

  • messages (Array<Any>) —

    An array of messages to chat with, e.g.

    ‘user’, content: ‘What is the best French cheese?’
  • model (String) (defaults to: nil) —

    The name of the model to chat with, e.g. mistral-tiny

  • tools (Array<Hash>) (defaults to: nil) —

    A list of tools to use.

  • temperature (Float) (defaults to: nil) —

    The temperature to use for sampling, e.g. 0.5.

  • max_tokens (Integer) (defaults to: nil) —

    The maximum number of tokens to generate, e.g. 100.

  • top_p (Float) (defaults to: nil) —

    The cumulative probability of tokens to generate, e.g. 0.9.

  • random_seed (Integer) (defaults to: nil) —

    The random seed to use for sampling, e.g. 42.

  • safe_mode (Boolean) (defaults to: false) —

    Deprecated, use safe_prompt instead.

  • safe_prompt (Boolean) (defaults to: false) —

    Whether to use safe prompt, e.g. true.

  • tool_choice (String, ToolChoice) (defaults to: nil) —

    The tool choice.

  • response_format (Hash<String, String>, ResponseFormat) (defaults to: nil) —

    The response format.

Returns:



95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
 
# File 'lib/mistral/client.rb', line 95

def chat_stream(
  messages:,
  model: nil,
  tools: nil,
  temperature: nil,
  max_tokens: nil,
  top_p: nil,
  random_seed: nil,
  safe_mode: false,
  safe_prompt: false,
  tool_choice: nil,
  response_format: nil
)
  request = make_chat_request(
    messages: messages,
    model: model,
    tools: tools,
    temperature: temperature,
    max_tokens: max_tokens,
    top_p: top_p,
    random_seed: random_seed,
    stream: true,
    safe_prompt: safe_mode || safe_prompt,
    tool_choice: tool_choice,
    response_format: response_format
  )

  Enumerator.new do |yielder|
    request('post', 'v1/chat/completions', json: request, stream: true).each do |json_response|
      yielder << ChatCompletionStreamResponse.new(**json_response)
    end
  end
end

#completion(model:, prompt:, suffix: nil, temperature: nil, max_tokens: nil, top_p: nil, random_seed: nil, stop: nil) ⇒ ChatCompletionResponse

A completion endpoint that returns a single response.

Parameters:

  • model (String) —

    model the name of the model to get completion with, e.g. codestral-latest

  • prompt (String) —

    the prompt to complete

  • suffix (String, nil) (defaults to: nil) —

    the suffix to append to the prompt for fill-in-the-middle completion

  • temperature (Float, nil) (defaults to: nil) —

    temperature the temperature to use for sampling, e.g. 0.5.

  • max_tokens (Integer, nil) (defaults to: nil) —

    the maximum number of tokens to generate, e.g. 100. Defaults to nil.

  • top_p (Float, nil) (defaults to: nil) —

    the cumulative probability of tokens to generate, e.g. 0.9. Defaults to nil.

  • random_seed (Integer, nil) (defaults to: nil) —

    the random seed to use for sampling, e.g. 42. Defaults to nil.

  • stop (Array<String>, nil) (defaults to: nil) —

    a list of tokens to stop generation at, e.g. [‘/n/n’]

Returns:

Raises:



173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
 
# File 'lib/mistral/client.rb', line 173

def completion(
  model:,
  prompt:,
  suffix: nil,
  temperature: nil,
  max_tokens: nil,
  top_p: nil,
  random_seed: nil,
  stop: nil
)
  request = make_completion_request(
    prompt:, model:, suffix:, temperature:, max_tokens:, top_p:, random_seed:, stop:
  )
  single_response = request('post', 'v1/fim/completions', json: request, stream: false)

  single_response.each do |response|
    return ChatCompletionResponse.new(**response)
  end

  raise Error, 'No response received'
end

#completion_stream(model:, prompt:, suffix: nil, temperature: nil, max_tokens: nil, top_p: nil, random_seed: nil, stop: nil) ⇒ Enumerator<ChatCompletionStreamResponse>

An asynchronous completion endpoint that streams responses.

Parameters:

  • model (String) —

    model the name of the model to get completions with, e.g. codestral-latest

  • prompt (String) —

    the prompt to complete

  • suffix (String, nil) (defaults to: nil) —

    the suffix to append to the prompt for fill-in-the-middle completion

  • temperature (Float, nil) (defaults to: nil) —

    temperature the temperature to use for sampling, e.g. 0.5.

  • max_tokens (Integer, nil) (defaults to: nil) —

    the maximum number of tokens to generate, e.g. 100. Defaults to nil.

  • top_p (Float, nil) (defaults to: nil) —

    the cumulative probability of tokens to generate, e.g. 0.9. Defaults to nil.

  • random_seed (Integer, nil) (defaults to: nil) —

    the random seed to use for sampling, e.g. 42. Defaults to nil.

  • stop (Array<String>, nil) (defaults to: nil) —

    a list of tokens to stop generation at, e.g. [‘/n/n’]

Returns:



208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
 
# File 'lib/mistral/client.rb', line 208

def completion_stream(
  model:,
  prompt:,
  suffix: nil,
  temperature: nil,
  max_tokens: nil,
  top_p: nil,
  random_seed: nil,
  stop: nil
)
  request = make_completion_request(
    prompt:, model:, suffix:, temperature:, max_tokens:, top_p:, random_seed:, stop:, stream: true
  )
  response = request('post', 'v1/fim/completions', json: request, stream: true)

  response.lazy.map do |json_streamed_response|
    ChatCompletionStreamResponse.new(**json_streamed_response)
  end
end

#embeddings(model:, input:) ⇒ EmbeddingResponse

An embeddings endpoint that returns embeddings for a single, or batch of inputs

Parameters:

  • model (String) —

    The embedding model to use, e.g. mistral-embed

  • input (String, Array<String>) —

    The input to embed, e.g. [‘What is the best French cheese?’]

Returns:

Raises:



136
137
138
139
140
141
142
143
144
145
 
# File 'lib/mistral/client.rb', line 136

def embeddings(model:, input:)
  request = { model: model, input: input }
  singleton_response = request('post', 'v1/embeddings', json: request)

  singleton_response.each do |response|
    return EmbeddingResponse.new(response)
  end

  raise Mistral::Error.new(message: 'No response received')
end

#list_models ⇒ ModelList

Returns a list of the available models

Returns:

  • (ModelList) —

    A response object containing the list of models.

Raises:



151
152
153
154
155
156
157
158
159
 
# File 'lib/mistral/client.rb', line 151

def list_models
  singleton_response = request('get', 'v1/models')

  singleton_response.each do |response|
    return ModelList.new(response)
  end

  raise Mistral::Error.new(message: 'No response received')
end