Openai.ex

Unofficial community-maintained wrapper for OpenAI REST APIs See https://platform.openai.com/docs/api-reference/introduction for further info on REST endpoints

⚠️⚠️⚠️Disclaimer: Please be advised that addressing issues or pull requests (PRs) may experience delays, as my current workload involves prioritizing other projects. Consequently, the library may not consistently reflect the latest API specifications. You can explore alternative projects here:

• https://github.com/restlessronin/openai_ex
• https://github.com/dvcrn/ex_openai

Thank you for your understanding and support and thanks to everyone who has contributed to the library so far!

Installation

Add :openai as a dependency in your mix.exs file.

def deps do
  [
    {:openai, "~> 0.6.2"}
  ]
end

Configuration

You can configure openai in your mix config.exs (default $project_root/config/config.exs). If you're using Phoenix add the configuration in your config/dev.exs|test.exs|prod.exs files. An example config is:

import Config

config :openai,
  # find it at https://platform.openai.com/account/api-keys
  api_key: "your-api-key",
  # find it at https://platform.openai.com/account/org-settings under "Organization ID"
  organization_key: "your-organization-key",
  # optional, use when required by an OpenAI API beta, e.g.:
  beta: "assistants=v1",
  # optional, passed to [HTTPoison.Request](https://hexdocs.pm/httpoison/HTTPoison.Request.html) options
  http_options: [recv_timeout: 30_000],
  # optional, useful if you want to do local integration tests using Bypass or similar
  # (https://github.com/PSPDFKit-labs/bypass), do not use it for production code,
  # but only in your test config!
  api_url: "http://localhost/"

Note: you can load your os ENV variables in the configuration file, if you set an env variable for API key named OPENAI_API_KEY you can get it in the code by doing System.get_env("OPENAI_API_KEY").

⚠️config.exs is compile time, so the get_env/1 function is executed during the build, if you want to get the env variables during runtime please use runtime.exs instead of config.exs in your application (elixir doc ref).

Configuration override

Client library configuration can be overwritten in runtime by passing a %OpenAI.Config{} struct as last argument of the function you need to use. For instance if you need to use a different api_key, organization_key or http_options you can simply do:

config_override = %OpenAI.Config{ api_key: "test-api-key" } # this will return a config struct with "test-api-key" as api_key, all the other config are defaulted by the client by using values taken from config.exs, so you don't need to set the defaults manually

# chat_completion with overriden config
OpenAI.chat_completion([
  model: "gpt-3.5-turbo",
  messages: [
        %{role: "system", content: "You are a helpful assistant."},
        %{role: "user", content: "Who won the world series in 2020?"},
        %{role: "assistant", content: "The Los Angeles Dodgers won the World Series in 2020."},
        %{role: "user", content: "Where was it played?"}
    ]
  ],
  config_override # <--- pass the overriden configuration as last argument of the function
)


# chat_completion with standard config
OpenAI.chat_completion(
  model: "gpt-3.5-turbo",
  messages: [
      %{role: "system", content: "You are a helpful assistant."},
      %{role: "user", content: "Who won the world series in 2020?"},
      %{role: "assistant", content: "The Los Angeles Dodgers won the World Series in 2020."},
      %{role: "user", content: "Where was it played?"}
  ]
)

you can perform a config override in all the functions, note that params argument must be passed explicitly as a list in square brackets if the configuration is to be overwritten, as in the example above.

Usage overview

Get your API key from https://platform.openai.com/account/api-keys

models()

Retrieve the list of available models

Example request

OpenAI.models()

Example response

{:ok, %{
  data: [%{
    "created" => 1651172505,
    "id" => "davinci-search-query",
    "object" => "model",
    "owned_by" => "openai-dev",
    "parent" => nil,
    "permission" => [
      %{
        "allow_create_engine" => false,
        "allow_fine_tuning" => false,
        "allow_logprobs" => true,
        ...
      }
    ],
    "root" => "davinci-search-query"
  },
  ....],
  object: "list"
}}

See: https://platform.openai.com/docs/api-reference/models/list

models(model_id)

Retrieve specific model info

OpenAI.models("davinci-search-query")

Example response

{:ok,
 %{
   created: 1651172505,
   id: "davinci-search-query",
   object: "model",
   owned_by: "openai-dev",
   parent: nil,
   permission: [
     %{
       "allow_create_engine" => false,
       "allow_fine_tuning" => false,
       "allow_logprobs" => true,
       "allow_sampling" => true,
       "allow_search_indices" => true,
       "allow_view" => true,
       "created" => 1669066353,
       "group" => nil,
       "id" => "modelperm-lYkiTZMmJMWm8jvkPx2duyHE",
       "is_blocking" => false,
       "object" => "model_permission",
       "organization" => "*"
     }
   ],
   root: "davinci-search-query"
 }}

See: https://platform.openai.com/docs/api-reference/models/retrieve

completions(params)

It returns one or more predicted completions given a prompt. The function accepts as arguments the "engine_id" and the set of parameters used by the Completions OpenAI api

Example request

  OpenAI.completions(
    model: "finetuned-model",
    prompt: "once upon a time",
    max_tokens: 5,
    temperature: 1,
    ...
  )

Example response

## Example response
  {:ok, %{
    choices: [
      %{
        "finish_reason" => "length",
        "index" => 0,
        "logprobs" => nil,
        "text" => "\" thing we are given"
      }
    ],
    created: 1617147958,
    id: "...",
    model: "...",
    object: "text_completion"
    }
  }

See: https://platform.openai.com/docs/api-reference/completions/create

completions(engine_id, params) (DEPRECATED)

this API has been deprecated by OpenAI, as engines are replaced by models. If you are using it consider to switch to completions(params) ASAP!

Example request

  OpenAI.completions(
    "davinci", # engine_id
    prompt: "once upon a time",
    max_tokens: 5,
    temperature: 1,
    ...
)

Example response

{:ok, %{
  choices: [
    %{
      "finish_reason" => "length",
      "index" => 0,
      "logprobs" => nil,
      "text" => "\" thing we are given"
    }
  ],
  created: 1617147958,
  id: "...",
  model: "...",
  object: "text_completion"
  }
}

See: https://beta.openai.com/docs/api-reference/completions/create for the complete list of parameters you can pass to the completions function

chat_completion()

Creates a completion for the chat message

Example request

OpenAI.chat_completion(
  model: "gpt-3.5-turbo",
  messages: [
        %{role: "system", content: "You are a helpful assistant."},
        %{role: "user", content: "Who won the world series in 2020?"},
        %{role: "assistant", content: "The Los Angeles Dodgers won the World Series in 2020."},
        %{role: "user", content: "Where was it played?"}
    ]
)

Example response

{:ok,
     %{
       choices: [
         %{
           "finish_reason" => "stop",
           "index" => 0,
           "message" => %{
             "content" =>
               "The 2020 World Series was played at Globe Life Field in Arlington, Texas due to the COVID-19 pandemic.",
             "role" => "assistant"
           }
         }
       ],
       created: 1_677_773_799,
       id: "chatcmpl-6pftfA4NO9pOQIdxao6Z4McDlx90l",
       model: "gpt-3.5-turbo-0301",
       object: "chat.completion",
       usage: %{
         "completion_tokens" => 26,
         "prompt_tokens" => 56,
         "total_tokens" => 82
       }
     }}

See: https://platform.openai.com/docs/api-reference/chat/create for the complete list of parameters you can pass to the completions function

chat_completion() with stream

Creates a completion for the chat message, by default it streams to self(), but you can override the configuration by passing a config override to the function with a different stream_to http_options parameter.

Example request

import Config

config :openai,
  api_key: "your-api-key",
  http_options: [recv_timeout: :infinity, async: :once],
  ...

http_options must be set as above when you want to treat the chat completion as a stream.

OpenAI.chat_completion([
    model: "gpt-3.5-turbo",
    messages: [
      %{role: "system", content: "You are a helpful assistant."},
      %{role: "user", content: "Who won the world series in 2020?"},
      %{role: "assistant", content: "The Los Angeles Dodgers won the World Series in 2020."},
      %{role: "user", content: "Where was it played?"}
    ],
    stream: true, # set this param to true
  ]
)
|> Stream.each(fn res ->
  IO.inspect(res)
end)
|> Stream.run()

Example response

%{
  "choices" => [
    %{"delta" => %{"role" => "assistant"}, "finish_reason" => nil, "index" => 0}
  ],
  "created" => 1682700668,
  "id" => "chatcmpl-7ALbIuLju70hXy3jPa3o5VVlrxR6a",
  "model" => "gpt-3.5-turbo-0301",
  "object" => "chat.completion.chunk"
}
%{
  "choices" => [
    %{"delta" => %{"content" => "The"}, "finish_reason" => nil, "index" => 0}
  ],
  "created" => 1682700668,
  "id" => "chatcmpl-7ALbIuLju70hXy3jPa3o5VVlrxR6a",
  "model" => "gpt-3.5-turbo-0