MultiCoT

Overview

MultiCoT is a repository that demonstrates Chain-of-Table reasoning on multiple tables powered by LangGraph.

Try the hosted demo: https://cyqiq-cot-demo-c27oq5sqda-uc.a.run.app/

Installation

Follow these steps to set up MultiCoT:

1. Clone the Repository

   git clone https://github.com/CYQIQ/MultiCoT.git
   

2. Install Dependencies Navigate to the repository directory and run:

   pip install -r requirements.txt
   

3. Configure API Keys Create a .env file in the root directory. Add your OpenAI API key and LangChain API details as follows:

   OPENAI_API_KEY="..."
   LANGCHAIN_API_KEY="..."
   LANGCHAIN_TRACING_V2="..."
   LANGCHAIN_ENDPOINT="..."
   LANGCHAIN_PROJECT="..."
   

Usage

For detailed instructions on how to run MultiCoT, refer to the CYQIQ_COT.ipynb Jupyter Notebook within the repository.

Data

The data directory includes an example dataset comprised of 6 CSV files necessary to run the notebook. To run MultiCoT with your own data, change the data loading logic and the table descriptions in CYQIQ_COT.ipynb

Help

Feel free to email us at [email protected]

Acknowledgments

• Big thanks to @hwchase17 and team LangChain for their amazing guides.
• Big thanks to @zaturalma2 for helping put the notebook together.