Skip to content

OpenAI

Sibila can use OpenAI remote models, for which you'll need a paid OpenAI account and its API key. Although you can pass this key when you create the model object, it's more secure to define an env variable with this information:

export OPENAI_API_KEY="..."

setx OPENAI_API_KEY "..."

Another possibility is to store your OpenAI key in .env files, which has many advantages: see the dotenv-python package.

Creating models#

OpenAI models can be used by Sibila through the OpenAIModel class.

Example

from sibila import OpenAIModel

model = OpenAIModel("gpt-3.5-turbo-0125")

model("I think that I shall never see.")

Result

'A poem as lovely as a tree.'

You can also create an OpenAI model in the Models factory by using the "openai:" provider prefix like this:

from sibila import Models

model = Models.create("openai:gpt-3.5-turbo-0125")

Model list#

The available OpenAI models are listed here. You can also fetch a list of known model names by calling OpenAIModel.known_models():

Example

OpenAIModel.known_models()

Result

['babbage-002',
 'dall-e-2',
 'dall-e-3',
 'davinci-002',
 'gpt-3.5-turbo',
 'gpt-3.5-turbo-0125',
 'gpt-3.5-turbo-0301',
 'gpt-3.5-turbo-0613',
 'gpt-3.5-turbo-1106',
 'gpt-3.5-turbo-16k',
 'gpt-3.5-turbo-16k-0613',
 'gpt-3.5-turbo-instruct',
 'gpt-3.5-turbo-instruct-0914',
 'gpt-4',
 'gpt-4-0125-preview',
 'gpt-4-0613',
 'gpt-4-1106-preview',
 'gpt-4-1106-vision-preview',
 'gpt-4-turbo-preview',
 'gpt-4-vision-preview',
 'text-embedding-3-large',
 'text-embedding-3-small',
 'text-embedding-ada-002',
 'tts-1',
 'tts-1-1106',
 'tts-1-hd',
 'tts-1-hd-1106',
 'whisper-1']

Not all of these models are for text inference, but the names that start with "gpt" are (excluding the "vision" models), and you can use those model names to create an OpenAI model.

JSON Schema models#

At the time of writing, not all OpenAI inference models support JSON Schema generation via the Tools functionality, which is required for structured data extraction. The following models (and later versions) allow JSON extraction:

  • gpt-3.5-turbo-1106 and later
  • gpt-4-1106-preview, gpt-4-turbo-preview and later

Using for other providers#

You can also use the OpenAIModel class to access any provider that uses the OpenAI API by setting the base_url and api_key arguments. For example to use the Together.ai service with the OpenAIModel class:

model_name = "mistralai/Mixtral-8x7B-Instruct-v0.1",

client = OpenAIModel(
    model_name,
    base_url="https://api.together.xyz/v1",
    api_key=os.environ["TOGETHER_API_KEY"],
)

This is just an example, as Together.ai has a dedicated Sibila class, but you can access any other OpenAI-compatible servers with the OpenAIModel class.