Codename Project IKARUS

This project is currently being worked on. More features and documentation will be added soon.

This project AIMS to deliver the experience utilizing langchain library with OpenAI model to do several useful tasks

• Question Answering from an external source of Knowledge base:
  • question_answering_from_knowledge_base.py script leverages the power of langchain and OpenAI inorder to perform question answering from provided pdf as knowledge sources.
  • Restricts knowledge only to the sources provided in the folder where the pdfs are stored.
  • Utilizes ongoing conversation in memory buffer to perform question answering.
• Natural Language to Structural Information
  • natural_language_to_structured_info.py script converts natural language text into a structured graph, linking entities to their Wikidata instances, thereby creating a knowledge base for future use as a prompt.
  • the general workflow is as follows:
    • take a free text as input
    • build a graph from the free text
    • connect the entities to the real wikidata instances
    • collect the connected entities from wikidata per entity
    • build a structured knowledge source in order to later use as a prompt for further usage.
• Question answering from structured file i.e, json.
  • question_answering_with_reasoning_from_KB.py script performs chain reasoning to answer questions about knowledge base stored in JSON file structure.

Installation

using pip

pip install -r requirements.txt

using conda

To begin with, one need to install the required software to operate the provided scripts. As a preliminary step, you must install the Conda environment, which you can find at https://docs.anaconda.com/anaconda/install/index.html. After installation using conda command line, you can create a new environment using the following command:

conda env create -f conda_env.yml

One need to activate the environment using the following command:

conda activate ikarus

Download the required spacy model using the following command:

python -m spacy download en_core_web_sm

Usage

To run Question Answering from an external source of Knowledge base:

python question_answering_from_knowledge_base.py --api_key <YOUR OPENAI API KEY> --directory_path <The directory of all the stored PDFs>

To run Natural Language to Structural Information

 python natural_language_to_structured_info.py --api_key <YOUR OPENAI API KEY> --text_or_filepath <The text (in natural language) or file_path>

Question answering from structured file i.e, json.

python question_answering_with_reasoning_from_KB.py --api_key <YOUR OPENAI API KEY> --file_path <The file path of the JSON file>