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

Installation

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

def deps do
  [
    {:openai, "~> 0.5.1"}
  ]
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, 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").

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
       }
     }}

Known issue: the stream param is not working properly in the current implementation

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

edits()

Creates a new edit for the provided input, instruction, and parameters

Example request

OpenAI.edits(
  model: "text-davinci-edit-001",
  input: "What day of the wek is it?",
  instruction: "Fix the spelling mistakes"
)

Example response

{:ok,
  %{
   choices: [%{"index" => 0, "text" => "What day of the week is it?\n"}],
   created: 1675443483,
   object: "edit",
   usage: %{
     "completion_tokens" => 28,
     "prompt_tokens" => 25,
     "total_tokens" => 53
  }
}}

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

images_generations(params)

This generates an image based on the given prompt. Image functions require some times to execute, and API may return a timeout error, if needed you can pass an optional configuration struct with HTTPoison http_options as second argument of the function to increase the timeout.

Example request

OpenAI.images_generations(
    [prompt: "A developer writing a test", size: "256x256"],
    %OpenAI.Config{http_options: [recv_timeout: 10 * 60 * 1000]} # optional!
 )

Example response

{:ok,
 %{
   created: 1670341737,
   data: [
     %{
       "url" => ...Returned url
     }
   ]
 }}

Note: this api signature has changed in v0.3.0 to be compliant with the conventions of other APIs, the alias OpenAI.image_generations(params, request_options) is still available for retrocompatibility. If you are using it consider to switch to OpenAI.images_generations(params, request_options) ASAP.

Note2: the official way of passing http_options changed in v0.5.0 to be compliant with the conventions of other APIs, the alias OpenAI.images_generations(file_path, params, request_options), but is still available for retrocompatibility. If you are using it consider to switch to OpenAI.images_variations(params, config)

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

images_edits(file_path, params)

Edit an existing image based on prompt Image functions require some times to execute, and API may return a timeout error, if needed you can pass an optional configuration struct with HTTPoison http_options as second argument of the function to increase the timeout.

Example Request

OpenAI.images_edits(
     "/home/developer/myImg.png",
     [prompt: "A developer writing a test", size: "256x256"],
    %OpenAI.Config{http_options: [recv_timeout: 10 * 60 * 1000]} # optional!
 )

Example Response

{:ok,
 %{
   created: 1670341737,
   data: [
     %{
       "url" => ...Returned url
     }
   ]
 }}

Note: the official way of passing http_options changed in v0.5.0 to be compliant with the conventions of other APIs, the alias OpenAI.images_edits(file_path, params, request_options), but is still available for retrocompatibility. If you are using it consider to switch to OpenAI.images_edits(file_path, params, config)

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

images_variations(file_path, params)

Image functions require some times to execute, and API may return a timeout error, if needed you can pass an optional configuration struct with HTTPoison http_options as second argument of the function to increase the timeout.

Example Request

OpenAI.images_variations(
    "/home/developer/myImg.png",
    [n: "5"],
    %OpenAI.Config{http_options: [recv_timeout: 10 * 60 * 1000]} # optional!
)

Example Response

{:ok,
 %{
   created: 1670341737,
   data: [
     %{
       "url" => ...Returned url
     }
   ]
 }}

Note: the official way of passing http_options changed in v0.5.0 to be compliant with the conventions of other APIs, the alias OpenAI.images_variations(file_path, params, request_options), but is still available for retrocompatibility. If you are using it consider to switch to OpenAI.images_edits(file_path, params, config)

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

embeddings(params)

Example request

OpenAI.embeddings(
    model: "text-embedding-ada-002",
    input: "The food was delicious and the waiter..."
  )

Example response

{:ok,
  %{
   data: [
     %{
       "embedding" => [0.0022523515000000003, -0.009276069000000001,
        0.015758524000000003, -0.007790373999999999, -0.004714223999999999,
        0.014806155000000001, -0.009803046499999999, -0.038323310000000006,
        -0.006844355, -0.028672641, 0.025345700000000002, 0.018145794000000003,
        -0.0035904291999999997, -0.025498080000000003, 5.142790000000001e-4,
        -0.016317246, 0.028444072, 0.0053713582, 0.009631619999999999,
        -0.016469626, -0.015390275, 0.004301531, 0.006984035499999999,
        -0.007079272499999999, -0.003926933, 0.018602932000000003, 0.008666554,
        -0.022717162999999995, 0.011460166999999997, 0.023860006,
        0.015568050999999998, -0.003587254600000001, -0.034843990000000005,
        -0.0041555012999999995, -0.026107594000000005, -0.02151083,
        -0.0057618289999999996, 0.011714132499999998, 0.008355445999999999,
        0.004098358999999999, 0.019199749999999998, -0.014336321, 0.008952264,
        0.0063395994, -0.04576447999999999, ...],
       "index" => 0,
       "object" => "embedding"
     }
   ],
   model: "text-embedding-ada-002-v2",
   object: "list",
   usage: %{"prompt_tokens" => 8, "total_tokens" => 8}
  }}

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

audio_transcription(file_path, params)

Transcribes audio into the input language.

Example request

OpenAI.audio_transcription(
  "./path_to_file/blade_runner.mp3", # file path
  model: "whisper-1"
)

Example response

 {:ok,
  %{
   text: "I've seen things you people wouldn't believe.."
  }}

See: https://platform.openai.com/docs/api-reference/audio/create to get info on the params accepted by the api

audio_translation(file_path, params)

Translates audio into into English.

Example request

OpenAI.audio_translation(
  "./path_to_file/werner_herzog_interview.mp3", # file path
  model: "whisper-1"
)

Example response

{:ok,
  %{
    text:  "I thought if I walked, I would be saved. It was almost like a pilgrimage. I will definitely continue to walk long distances. It is a very unique form of life and existence that we have lost almost entirely from our normal life."
  }
}

See: https://platform.openai.com/docs/api-reference/audio/create to get info on the params accepted by the api

files()

Returns a list of files that belong to the user's organization.

Example request

OpenAI.files()

Example response

{:ok,
 %{
 data: [
   %{
     "bytes" => 123,
     "created_at" => 213,
     "filename" => "file.jsonl",
     "id" => "file-123321",
     "object" => "file",
     "purpose" => "fine-tune",
     "status" => "processed",
     "status_details" => nil
   }
 ],
 object: "list"
 }
}

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

files(file_id)

Returns a file that belong to the user's organization, given a file id

Example request

OpenAI.files("file-123321")

Example response

{:ok,
%{
  bytes: 923,
  created_at: 1675370979,
  filename: "file.jsonl",
  id: "file-123321",
  object: "file",
  purpose: "fine-tune",
  status: "processed",
  status_details: nil
}
}

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

files_upload(file_path, params)

Upload a file that contains document(s) to be used across various endpoints/features. Currently, the size of all the files uploaded by one organization can be up to 1 GB. Please contact OpenAI if you need to increase the storage limit.

Example request

OpenAI.files_upload("./file.jsonl", purpose: "fine-tune")

Example response

{:ok,
  %{
    bytes: 923,
    created_at: 1675373519,
    filename: "file.jsonl",
    id: "file-123",
    object: "file",
    purpose: "fine-tune",
    status: "uploaded",
    status_details: nil
  }
}

See: https://platform.openai.com/docs/api-reference/files/upload

files_delete(file_id)

delete a file

Example request

OpenAI.files_delete("file-123")

Example response

{:ok, %{deleted: true, id: "file-123", object: "file"}}

See: https://platform.openai.com/docs/api-reference/files/delete

finetunes()

List your organization's fine-tuning jobs.

Example request

OpenAI.finetunes()

Example response

{:ok,
  %{
    object: "list",
    data: [%{
      "id" => "t-AF1WoRqd3aJAHsqc9NY7iL8F",
      "object" => "fine-tune",
      "model" => "curie",
      "created_at" => 1614807352,
      "fine_tuned_model" => null,
      "hyperparams" => { ... },
      "organization_id" => "org-...",
      "result_files" = [],
      "status": "pending",
      "validation_files" => [],
      "training_files" => [ { ... } ],
      "updated_at" => 1614807352,
    }],
  }
}

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

finetunes(finetune_id)

Gets info about a fine-tune job.

Example request

OpenAI.finetunes("t-AF1WoRqd3aJAHsqc9NY7iL8F")

Example response

{:ok,
  %{
    object: "list",
    data: [%{
      "id" => "t-AF1WoRqd3aJAHsqc9NY7iL8F",
      "object" => "fine-tune",
      "model" => "curie",
      "created_at" => 1614807352,
      "fine_tuned_model" => null,
      "hyperparams" => { ... },
      "organization_id" => "org-...",
      "result_files" = [],
      "status": "pending",
      "validation_files" => [],
      "training_files" => [ { ... } ],
      "updated_at" => 1614807352,
    }],
  }
}

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

finetunes_create(params)

Creates a job that fine-tunes a specified model from a given dataset.

Example request

OpenAI.finetunes_create(
  training_file: "file-123213231",
  model: "curie",
)

Example response

{:ok,                                                                           
 %{                                                                             
   created_at: 1675527767,                                                      
   events: [                                                                    
     %{                                                                         
       "created_at" => 1675527767,                                              
       "level" => "info",                                                       
       "message" => "Created fine-tune: ft-IaBYfSSAK47UUCbebY5tBIEj",           
       "object" => "fine-tune-event"                                            
     }                                                                          
   ],                                                                           
   fine_tuned_model: nil,                                                       
   hyperparams: %{                                                              
     "batch_size" => nil,                                                       
     "learning_rate_multiplier" => nil,                                         
     "n_epochs" => 4,                                                           
     "prompt_loss_weight" => 0.01                                               
   },                                                                           
   id: "ft-IaBYfSSAK47UUCbebY5tBIEj",                                           
   model: "curie",                                                              
   object: "fine-tune",                                                         
   organization_id: "org-1iPTOIak4b5fpuIB697AYMmO",                             
   result_files: [],                                                            
   status: "pending",                                                           
   training_files: [                                                            
     %{                                                                         
       "bytes" => 923,                                                          
       "created_at" => 1675373519,                                              
       "filename" => "file-12321323.jsonl",                                             
       "id" => "file-12321323",                                 
       "object" => "file",                                                      
       "purpose" => "fine-tune",                                                
       "status" => "processed",                                                 
       "status_details" => nil                                                  
     }                                                                          
   ],                                                                           
   updated_at: 1675527767,                                                      
   validation_files: []                                                         
 }}

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

finetunes_list_events(finetune_id)

Get fine-grained status updates for a fine-tune job.

Example request

OpenAI.finetunes_list_events("ft-AF1WoRqd3aJAHsqc9NY7iL8F")

Example response

{:ok,
  %{
   data: [
     %{
       "created_at" => 1675376995,
       "level" => "info",
       "message" => "Created fine-tune: ft-123",
       "object" => "fine-tune-event"
     },
     %{
       "created_at" => 1675377104,
       "level" => "info",
       "message" => "Fine-tune costs $0.00",
       "object" => "fine-tune-event"
     },
     %{
       "created_at" => 1675377105,
       "level" => "info",
       "message" => "Fine-tune enqueued. Queue number: 18",
       "object" => "fine-tune-event"
     },
    ...,
     ]
    }
  }

See: https://platform.openai.com/docs/api-reference/fine-tunes/events

finetunes_cancel(finetune_id)

Immediately cancel a fine-tune job.

Example request

OpenAI.finetunes_cancel("ft-AF1WoRqd3aJAHsqc9NY7iL8F")

Example response

  {:ok,
  %{
   created_at: 1675527767,
   events: [
     ...
     %{
       "created_at" => 1675528080,
       "level" => "info",
       "message" => "Fine-tune cancelled",
       "object" => "fine-tune-event"
     }
   ],
   fine_tuned_model: nil,
   hyperparams: %{
     "batch_size" => 1,
     "learning_rate_multiplier" => 0.1,
     "n_epochs" => 4,
     "prompt_loss_weight" => 0.01
   },
   id: "ft-IaBYfSSAK47UUCbebY5tBIEj",
   model: "curie",
   object: "fine-tune",
   organization_id: "org-1iPTOIak4b5fpuIB697AYMmO",
   result_files: [],
   status: "cancelled",
   training_files: [
     %{
       "bytes" => 923,
       "created_at" => 1675373519,
       "filename" => "file123.jsonl",
       "id" => "file-123",
       "object" => "file",
       "purpose" => "fine-tune",
       "status" => "processed",
       "status_details" => nil
     }
   ],
   updated_at: 1675528080,
   validation_files: []
  }}

finetunes_delete_model(finetune_id)

Immediately cancel a fine-tune job.

Example request

OpenAI.finetunes_delete_model("model-id")

Example response

{:ok,
  %{
   id: "model-id",
   object: "model",
   deleted: true
  }
}

See: https://platform.openai.com/docs/api-reference/fine-tunes/delete-model

moderations(params)

Classifies if text violates OpenAI's Content Policy

Example request

OpenAI.moderations(input: "I want to kill everyone!")

Example response

{:ok,
  %{
   id: "modr-6gEWXyuaU8dqiHpbAHIsdru0zuC88",
   model: "text-moderation-004",
   results: [
     %{
       "categories" => %{
         "hate" => false,
         "hate/threatening" => false,
         "self-harm" => false,
         "sexual" => false,
         "sexual/minors" => false,
         "violence" => true,
         "violence/graphic" => false
       },
       "category_scores" => %{
         "hate" => 0.05119025334715844,
         "hate/threatening" => 0.00321022979915142,
         "self-harm" => 7.337320857914165e-5,
         "sexual" => 1.1111642379546538e-6,
         "sexual/minors" => 3.588798147546868e-10,
         "violence" => 0.9190407395362855,
         "violence/graphic" => 1.2791929293598514e-7
       },
       "flagged" => true
     }
   ]
  }}

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

Deprecated APIs

The following APIs are deprecated, but currently supported by the library for retrocompatibility with older versions. If you are using the following APIs consider to remove it ASAP from your project!

Note: from version 0.5.0 search, answers, classifications API are not supported (since they has been removed by OpenAI), if you still need them consider to use v0.4.2

engines() (DEPRECATED: use models instead)

Get the list of available engines

Example request

OpenAI.engines()

Example response

{:ok, %{
  "data" => [
    %{"id" => "davinci", "object" => "engine", "max_replicas": ...},
    ...,
    ...
  ]
}

See: https://beta.openai.com/docs/api-reference/engines/list

engines(engine_id)

Retrieve specific engine info

Example request

OpenAI.engines("davinci")

Example response

{:ok, %{
    "id" => "davinci",
    "object" => "engine",
    "max_replicas": ...
  }
}

See: https://beta.openai.com/docs/api-reference/engines/retrieve

TODO

improve JSON decoding strategy and performance #13
add support to API streaming (SSE)

License

The package is available as open source under the terms of the MIT License.