Thematic Structurizer is a Python package designed to process user-provided text related to historical or thematic content—such as summaries, descriptions, or analyses of topics like persuasion techniques from antiquity—and extract structured, pattern-validated data. It leverages large language models (LLMs) to generate responses in a predefined format, ensuring consistency and reliability through regex validation and retry mechanisms.

Install the package via pip:

pip install thematic_structurizer

Here's an example of how to use the package:

from thematic_structurizer import thematic_structurizer

# Example user input
user_input = "Describe the persuasion techniques used by Cicero in ancient Rome."

# Calling the function with default LLM (ChatLLM7)
response = thematic_structurizer(user_input)
print(response)

The package uses ChatLLM7 from langchain_llm7 by default. You can provide your own language model instance to customize the behavior. Supported models include those from OpenAI, Anthropic, Google Generative AI, etc.

from langchain_openai import ChatOpenAI
from thematic_structurizer import thematic_structurizer

llm = ChatOpenAI()
response = thematic_structurizer(user_input, llm=llm)

from langchain_anthropic import ChatAnthropic
from thematic_structurizer import thematic_structurizer

llm = ChatAnthropic()
response = thematic_structurizer(user_input, llm=llm)

For the default ChatLLM7, you can set your API key via environment variable:

export LLM7_API_KEY='your_api_key'

or pass it directly:

response = thematic_structurizer(user_input, api_key='your_api_key')

To obtain a free API key, register at https://token.llm7.io/.

def thematic_structurizer(
    user_input: str,
    api_key: Optional[str] = None,
    llm: Optional[BaseChatModel] = None
) -> List[str]:

• user_input: The text to process, such as a description, summary, or analysis.
• llm: An optional language model instance; defaults to ChatLLM7.
• api_key: Optional string; API key for ChatLLM7.

This function processes the input, invokes the LLM, and returns a list of extracted data that match the predefined pattern validation.

This project is maintained by Eugene Evstafev. For issues or contributions, please visit https://github.com/chigwell/thematic-structurizer.

• Email: hi@eugene.plus
• GitHub: chigwell