OpenAIAutomataAgentConfig.Config

The OpenAIAutomataAgentConfig.Config class is a configuration class that is used within OpenAIAutomataAgentConfig. It manages configurations related to OpenAI models and set up parameters for the Automata agent such as which OpenAI engine models are supported and whether arbitrary types are allowed.

The computation setup and the model are crucial elements while building an agent. This class provides configurations for such important elements enabling developers to build and utilize the agents effectively.

Overview

The class includes attributes such as SUPPORTED_MODELS and arbitrary_types_allowed. The SUPPORTED_MODELS attribute is a list of OpenAI engine model names that identifies which models are supported to help ensure compatibility and performance. The arbitrary_types_allowed attribute dictates whether arbitrary types are allowed in the model.

Example

Here is an example of how you might build and use an OpenAIAutomataAgentConfig with a specified model:

from automata.config.openai_agent import OpenAIAutomataAgentConfig
from automata.config.base import AgentConfigName

config = OpenAIAutomataAgentConfig.Config()
config_name = AgentConfigName.AUTOMATA_MAIN # get config name
config.model = "gpt-4"  # Set model
config.arbitrary_types_allowed = True

Limitations

OpenAIAutomataAgentConfig.Config can only use the predefined models listed in SUPPORTED_MODELS. If you need to use a different model, you would need to update the SUPPORTED_MODELS attribute in the class.

Follow-up Questions:

Currently, the class uses a hard-coded list for supported models. Can it potentially support all models in the future?
How does the class handle arbitrary types in a safe and performant manner?