script.py

import argparse
import os
import sys

from datetime import datetime

from app.content_handler import parse_content, save_in_disk
from app.services.open_ai import create_completion, upload_file

RAW_DOCUMENTS_PATH = "./documents"
DESTINATION_DIRECTORY = "output"

if not (len(sys.argv) > 1):
    raise ValueError("Please provide a question to be answered.")

parser = argparse.ArgumentParser()
parser.add_argument("--question", help="Specify the question to be answered")
args = parser.parse_args()

QUESTION = args.question

if __name__ == "__main__":
    all_items = os.listdir(RAW_DOCUMENTS_PATH)
    files = [{"path": f"{RAW_DOCUMENTS_PATH}/{item}", "name": os.path.splitext(item)[0]} for item in all_items]

    # Parse the content of each file and save it in a txt file
    for file in files:
        content = parse_content(file["path"])
        file["txt_output_path"] = save_in_disk(DESTINATION_DIRECTORY, file["name"], content)

    # Upload the files to OpenAI
    context = ""
    for file in files:
        file["upload_id"] = upload_file(file["txt_output_path"])

    context = ""
    for file in files:
        context += f"- {file['name']} (File ID: {file['upload_id']})\n"

    prompt = f"""
        Analyze the following documents based on the uploaded files, and than answer the question giving examples:

        Documents:
        {context}

        ---
        Question:
        {QUESTION}
    """

    completion = create_completion(prompt)

    report_name = f'ai_report_{datetime.now()}'.replace(' ', '_')
    save_in_disk(DESTINATION_DIRECTORY, report_name, completion, "md")

    print(f"Report saved to {report_name}")