Accompanying code to go with a series of posts on Retrieval Augmented Generation posts on LinkedIn by Harsh Singhal https://www.linkedin.com/in/harshsinghal/

query_llm.py

#          ___         
#         / ()\\        
#       _|_____|_       
#      | | === | |      
#      |_|  O  |_|        
#       ||  O  ||         
#       ||__*__||         
#      |~ \\___/ ~|       
#      /=\\ /=\\ /=\\     
#______[_]_[_]_[_]_______


import argparse
from llama_index import (
    GPTVectorStoreIndex,
    SimpleDirectoryReader,
    ServiceContext,
    StorageContext,
    LLMPredictor,
    load_index_from_storage,
)
from langchain.chat_models import ChatOpenAI
import openai
import os

def main():
    # Initialize the argument parser
    parser = argparse.ArgumentParser(description='Query the LLM with a given question.')

    # Add the question argument to the parser. This will allow users to input their question as a command-line argument.
    parser.add_argument('question', type=str, help='The question you want to ask the LLM.')

    # Parse the arguments provided by the user
    args = parser.parse_args()

    # Set the OpenAI API key for authentication.
    # NOTE: Ensure to keep your API keys private and secure. Avoid hardcoding them in scripts.
    os.environ['OPENAI_API_KEY'] = 'YOUR_API_KEY'
    openai.api_key = 'YOUR_API_KEY'
    
    # Define paths for the documents and index. Adjust these paths according to your directory structure.
    documents_folder = "./aws_case_documents/"
    index_name = "./aws_case_documents_index"

    # Initialize the LLM predictor with the desired model and temperature setting.
    llm_predictor = LLMPredictor(llm=ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0))
    
    # Setup the service context using the initialized LLM predictor.
    service_context = ServiceContext.from_defaults(llm_predictor=llm_predictor)

    # Define the storage context, pointing it to the directory where the index is stored.
    storage_context = StorageContext.from_defaults(persist_dir=index_name)

    # Load the previously built index from storage. This index will be used to query the documents.
    index = load_index_from_storage(storage_context=storage_context, service_context=service_context)

    # Use the index's query engine to query the provided question.
    response = index.as_query_engine().query(args.question)

    # Print the response to the console.
    print(response)

# This conditional checks if the script is being run as a standalone file and not being imported as a module.
if __name__ == '__main__':
    main()

• Replace 'YOUR_API_KEY' with their actual OpenAI API key.
• Install the required libraries/packages.
• Adjust the documents_folder and index_name paths according to their directory structure.

Run the script using:

python script_name.py "Your question here"