Idealist

Idealist is a flexible Python library that leverages Large Language Models (LLMs) and embeddings to generate and refine creative ideas while avoiding repetition. It's ideal for generating unique names, themes, or any other creative content.

Features

• Dynamic Idea Generation: Customize parameters to generate a wide range of ideas.
• Model Flexibility: Use different LLMs and embedding models supported by LiteLLM.
• Uniqueness Assurance: Ensures all generated ideas are unique and not duplicated.
• Embedding Support: Uses embeddings to find and avoid similar ideas.
• Easy to Use: Simple interface with minimal setup.

Installation

You can install the library via pip:

pip install infinite-idealist

Usage

Create a New Generator

Here's a basic example of how to use the IdeaGenerator class to create a new generator:

from idea_generator import IdeaGenerator
import os

def main():
    # Set up environment variables for API keys (optional, depending on the model used)
    os.environ["OPENAI_API_KEY"] = "your_openai_api_key"
    # os.environ["ANTHROPIC_API_KEY"] = "your_anthropic_api_key"  # Optional if using Anthropic

    # Initialize the IdeaGenerator
    generator = IdeaGenerator(
        name="Pokemon Names",
        description="Generate unique and creative names for Pokemon characters",
        model="gpt-4o-mini",  # Specify the LLM model
        embedding_model="text-embedding-ada-002",  # Specify the embedding model
        openai_api_key=os.getenv("OPENAI_API_KEY"),  # Optional if using OpenAI
        anthropic_api_key=os.getenv("ANTHROPIC_API_KEY"),  # Optional if using Anthropic
        max_recent_ideas=15,  # Number of recent ideas to include in prompts
        debug=False  # Set to True for detailed logs
    )

    # Setup parameters
    parameters = {
        "name": "Name of Pokemon"
    }

    generator.setup_parameters(parameters)

    # Generate ideas
    for i in range(5):
        idea = generator.generate_idea()
        if idea:
            print(f"Idea #{i+1}: {idea.get('name')}")
        else:
            print(f"Failed to generate Idea #{i+1}")

if __name__ == "__main__":
    main()

Load an Existing Generator

You can also load an existing generator by its ID:

from idea_generator import IdeaGenerator
import os

def main():
    # Set up environment variables for API keys (optional, depending on the model used)
    os.environ["OPENAI_API_KEY"] = "your_openai_api_key"
    # os.environ["ANTHROPIC_API_KEY"] = "your_anthropic_api_key"  # Optional if using Anthropic

    # Load an existing IdeaGenerator by ID
    generator_id = "pokemon_names_20241212_050458"  # Replace with your generator ID
    generator = IdeaGenerator.load(
        generator_id=generator_id,
        model="gpt-4o-mini",  # Specify the LLM model used during creation
        embedding_model="text-embedding-ada-002",  # Specify the embedding model used during creation
        openai_api_key=os.getenv("OPENAI_API_KEY"),  # Optional if using OpenAI
        anthropic_api_key=os.getenv("ANTHROPIC_API_KEY"),  # Optional if using Anthropic
        max_recent_ideas=15,  # Number of recent ideas to include in prompts
        debug=False  # Set to True for detailed logs
    )

    # Generate ideas
    for i in range(5):
        idea = generator.generate_idea()
        if idea:
            print(f"Idea #{i+1}: {idea.get('name')}")
        else:
            print(f"Failed to generate Idea #{i+1}")

if __name__ == "__main__":
    main()

API Reference

IdeaGenerator Class

__init__

IdeaGenerator(
    name: str,
    description: str,
    model: str = "gpt-4o-mini",
    embedding_model: str = "text-embedding-ada-002",
    anthropic_api_key: Optional[str] = None,
    openai_api_key: Optional[str] = None,
    max_recent_ideas: int = 20,
    debug: bool = False,
    generator_id: Optional[str] = None
)

• name: Name of the generator.
• description: Description of what this generator creates.
• model: The LLM model to use (e.g., 'gpt-4o-mini').
• embedding_model: The embedding model to use (e.g., 'text-embedding-ada-002').
• anthropic_api_key: API key for Anthropic (optional).
• openai_api_key: API key for OpenAI (optional).
• max_recent_ideas: Maximum number of recent ideas to include in prompts.
• debug: Enables detailed logging if set to True.
• generator_id: Optional ID to load an existing generator.

load Class Method

@classmethod
load(
    generator_id: str,
    model: str = "gpt-4o-mini",
    embedding_model: str = "text-embedding-ada-002",
    anthropic_api_key: Optional[str] = None,
    openai_api_key: Optional[str] = None,
    max_recent_ideas: int = 20,
    debug: bool = False
) -> 'IdeaGenerator'

• generator_id: ID of the generator to load.
• model: The LLM model to use (must match the one used during creation).
• embedding_model: The embedding model to use (must match the one used during creation).
• anthropic_api_key: API key for Anthropic (optional).
• openai_api_key: API key for OpenAI (optional).
• max_recent_ideas: Maximum number of recent ideas to include in prompts.
• debug: Enables detailed logging if set to True.

setup_parameters

setup_parameters(parameters: Dict[str, Any])

• parameters: Dictionary defining the structure of generated ideas.

generate_idea

generate_idea() -> Dict

• Generates a new idea based on the configured parameters.
• Returns a dictionary containing the generated idea.

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.

The GitHub repository for this project can be found at: https://github.com/yoheinakajima/idealist

License

This project is licensed under the MIT License. See the LICENSE file for details.