AutoGen Data Interpreter (DI)

Implemented in autoGen's GroupChat framework with reference to metaGPT's Data Interpreter module, with minor modifications to include tool recommendation and experience pool retrieval.

1. Environment Setup

   • Create a .env file in the root directory to store the API key.
   • Add the following content to the file:

     OPENAI_API_KEY=sk-xxxxxxxx
     

   • Install the required Python libraries:

     pip install python-dotenv requests faiss numpy

2. Functionality Overview

   • get_embedding(text, model): This function retrieves embeddings for a given text using a specified model.
   • search_answer(question, index, questions, answers): This function searches for answers based on a given question using Faiss indexing.
   • Other functions and variable descriptions can be found within the codebase.

3. Code Execution

   • The DI agent interacts with the user to understand data-related queries.
   • It formulates a plan based on the user's requirements.
   • The agent then proceeds to write code to address the problem at hand.
   • Tools are utilized as needed to enhance the data analysis process.

4. Technical Details

   • For a comprehensive understanding of the design and technical aspects of the Data Interpreter, please refer to our research paper:
     • Data Interpreter: A Comprehensive Tool for Data Analysis and Processing

5. Author

   • mcxiaoxiao