Python API

OpenAI

class RPA.OpenAI.OpenAI

Library to support OpenAI and Azure OpenAI services.

Library is not included in the rpaframework package, so in order to use it you have to add rpaframework-openai with the desired version in your conda.yaml file.

Robot Framework example usage

*** Settings ***
Library    RPA.Robocorp.Vault
Library    RPA.OpenAI

*** Tasks ***
Create a text completion
    ${secrets}   Get Secret   secret_name=OpenAI
    Authorize To OpenAI   api_key=${secrets}[key]
    ${completion}    Completion Create
    ...     Write a tagline for an ice cream shop
    ...     temperature=0.6
    Log   ${completion}

Python example usage

from RPA.Robocorp.Vault import Vault
from RPA.OpenAI import OpenAI

secrets = Vault().get_secret("OpenAI")
baselib = OpenAI()
baselib.authorize_to_openai(secrets["key"])

result = baselib.completion_create(
    Create a tagline for icecream shop',
    temperature=0.6,
)
print(result)

ROBOT_LIBRARY_DOC_FORMAT = 'REST'

ROBOT_LIBRARY_SCOPE = 'GLOBAL'

authorize_to_azure_openai(api_key: str, api_base: str, api_type: Optional[str] = 'azure', api_version: Optional[str] = '2023-05-15') → None

Keyword for authorize to Azure OpenAI.

Parameters

api_key – Your Azure OpenAI API key
api_base – Your Endpoint URL. Example: https://docs-test-001.openai.azure.com/
api_type – “azure”
api_version – “2023-05-15”

Robot Framework example:

${secrets}   Get Secret   secret_name=AzureOpenAI
Authorize To Azure Openai
...    api_key=${secrets}[api_key]
...    api_base=${secrets}[api_base]
...    api_type=azure
...    api_version=2023-05-15

Python example:

secrets = Vault().get_secret("AzureOpenAI")
baselib = OpenAI()
baselib.authorize_to_azure_openai(
    secrets["api_key"],
    secrets["api_base"],
    "azure",
    "2023-05-15"
)

authorize_to_openai(api_key: str) → None

Keyword for authorize to OpenAI with your API key obtained from your account.

Parameters: api_key – Your OpenAI API key

Robot Framework example:

${secrets}   Get Secret   secret_name=OpenAI
Authorize To OpenAI   api_key=${secrets}[key]

Python example:

secrets = Vault().get_secret("OpenAI")
baselib = OpenAI()
baselib.authorize_to_openai(secrets["key"])

chat_completion_create(user_content: Optional[str] = None, conversation: Optional[List] = None, model: Optional[str] = 'gpt-3.5-turbo', system_content: Optional[str] = None, temperature: Optional[int] = 1, top_probability: Optional[int] = 1, frequency_penalty: Optional[int] = 0, presence_penalty: Optional[int] = 0) → None

Keyword for creating ChatGPT text completions using OpenAI or Azure OpenAI. Keyword returns the response as a string and the message history as a list.

Note. When using Azure OpenAI you must provide the deployment_name as the model parameter instead of the model ID used with OpenAI.

Parameters

user_content – Text submitted to ChatGPT to generate completions.
conversation – List containing the conversation to be continued. Leave empty for a new conversation.
model – For OpenAI the ID of the model to use, e.g. gpt-4 or gpt-3.5-turbo. For Azure OpenAI the Deployment name, e.g. myGPT4deployment.
system_content – The system message helps set the behavior of the assistant.
temperature – What sampling temperature to use between 0 to 2. Higher values means the model will take more risks.
top_probability – An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass.
frequency_penalty – Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far.
presence_penalty – Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far.

Robot Framework example:

# Get response without conversation history.
${response}   @{chatgpt_conversation}=     Chat Completion Create
...    user_content=What is the biggest mammal?
Log    ${response}

# Continue the conversation by using the "conversation" argument.
${response}   @{chatgpt_conversation}=     Chat Completion Create
...    conversation=${chatgpt_conversation}
...    user_content=How old can it live?
Log    ${response}

completion_create(prompt: str, model: Optional[str] = 'text-davinci-003', temperature: Optional[int] = 0.7, max_tokens: Optional[int] = 256, top_probability: Optional[int] = 1, frequency_penalty: Optional[int] = 0, presence_penalty: Optional[int] = 0, result_format: Optional[str] = 'string') → None

Keyword for creating text completions in OpenAI and Azure OpenAI. Keyword returns a text string.

Note. When using Azure OpenAI you must provide the deployment_name as the model parameter instead of the model ID used with OpenAI.

Parameters

prompt – Text submitted to OpenAI for creating natural language.
model – For OpenAI the ID of the model to use, e.g. text-davinci-003. For Azure OpenAI the Deployment name, e.g. myDavinci3deployment.
temperature – What sampling temperature to use. Higher values means the model will take more risks..
max_tokens – The maximum number of tokens to generate in the completion..
top_probability – Controls diversity via nucleus sampling. 0.5 means half of all likelihood-weighted options are considered.
frequency_penalty – Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far.
presence_penalty – Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far.
result_format – Result format (string / json). Return just a string or the default JSON response.

Robot Framework example:

${response}  Completion Create
...     Write a tagline for an icecream shop.
...     temperature=0.6
Log     ${response}

Python example:

result = baselib.completion_create(
    'Create a tagline for icecream shop',
    temperature=0.6,
)
print(result)

image_create(prompt: str, size: Optional[str] = '512x512', num_images: Optional[int] = 1, result_format: Optional[str] = 'list') → None

Keyword for creating one or more images using OpenAI. Keyword returns a list of urls for the images created.

Note. Keyword not supported in the Azure OpenAI service.

Parameters

prompt – A text description of the desired image(s). The maximum length is 1000 characters.
size – Size of the files to be created. 256x256, 512x512, 1024x1024
num_images – The number of images to generate. Must be between 1 and 10.
result_format – Result format (list / json).

Robot Framework example:

${images}    Image Create
...   Cartoon style picture of a cute monkey skateboarding.
...   size=256x256
...   num_images=2
FOR    ${url}    IN    @{images}
    Log    ${url}
END

Python example:

images = baselib.image_create(
    'Cartoon style picture of a cute monkey skateboarding',
    size='256x256',
    num_images=2,
)
for url in images:
    print(url)

image_create_variation(src_image: str, size: Optional[str] = '512x512', num_images: Optional[int] = 1, result_format: Optional[str] = 'list') → None

Keyword for creating one or more variations of a image. Keyword returns a list of urls for the images created. Source file must be a valid PNG file, less than 4MB, and square.

Note. Keyword not supported in the Azure OpenAI service.

Parameters

src_image – The image to use as the basis for the variation(s). Must be a valid PNG file, less than 4MB, and square.
size – The size of the generated images. Must be one of 256x256, 512x512, or 1024x1024.
num_images – The number of images to generate. Must be between 1 and 10
result_format – Result format (list / json).

Robot Framework example:

${variations}   Image Create Variation
...     source_image.png
...     size=256x256
...     num_images=2
FOR    ${url}    IN    @{variations}
    Log    ${url}
END

Python example:

variations = baselib.image_create_variation(
    'source_image.png',
    size='256x256',
    num_images=2,
)
for url in variations:
    print(url)