A text-based analysis package inspired by Nassim Taleb's concepts of Black Swan events and antifragility.
A new package would provide a text input describing a system, event, or scenario, and return a structured analysis based on Nassim Taleb's concepts of Black Swan events and antifragility. It would identify whether the input describes a fragile, robust, or antifragile system, highlight potential hidden risks (Black Swans), and suggest principles to improve resilience or benefit from volatility.
You can install the package using pip:
pip install taleb_analysisTo use the package, you can call the taleb_analysis function with the following parameters:
from taleb_analysis import taleb_analysis
value = taleb_analysis(user_input="text to analyze", api_key=None, llm=None)By default, it will use the ChatLLM7 from langchain_llm7 package as the LLM instance. You can pass your own llm instance if you prefer to use a different LLM.
For example, to use the OpenAI's LLM, you can pass it like this:
from langchain_openai import ChatOpenAI
from taleb_analysis import taleb_analysis
llm = ChatOpenAI()
response = taleb_analysis(user_input="text to analyze", llm=llm)Similarly, for Anthenropic:
from langchain_anthropic import ChatAnthropic
from taleb_analysis import taleb_analysis
llm = ChatAnthropic()
response = taleb_analysis(user_input="text to analyze", llm=llm)And for Google Generative AI:
from langchain_google_genai import ChatGoogleGenerativeAI
from taleb_analysis import taleb_analysis
llm = ChatGoogleGenerativeAI()
response = taleb_analysis(user_input="text to analyze", llm=llm)The default rate limits for the ChatLLM7 free tier are generally sufficient. For higher rate limits, you can provide your API key in one of the following ways:
- Set the
LLM7_API_KEYenvironment variable. - Pass the API key directly to the function:
taleb_analysis(user_input, api_key="your_api_key").
You can obtain a free API key by registering at https://token.llm7.io/
Contributions are welcome! Please refer to the GitHub repository for more information.
- Eugene Evstafev (hi@eugene.plus)