https://github.com/obaskly/docai

GPT-3 based Question Answering System that reads text from PDF, DOCX, or TXT files and answers questions based on the content.
https://github.com/obaskly/docai

ai chatbot chatgpt gpt gui python

Last synced: 7 days ago
JSON representation

GPT-3 based Question Answering System that reads text from PDF, DOCX, or TXT files and answers questions based on the content.

• Host: GitHub
• URL: https://github.com/obaskly/docai
• Owner: obaskly
• Created: 2023-04-08T23:04:37.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2023-04-09T02:16:31.000Z (almost 2 years ago)
• Last Synced: 2024-11-11T01:35:24.467Z (2 months ago)
• Topics: ai, chatbot, chatgpt, gpt, gui, python
• Language: Python
• Homepage:
• Size: 22.5 KB
• Stars: 4
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        



  


  

  


  Docai

  




Docai is a GPT-3 based Question Answering System that can provide answers based on a PDF, DOCX, and TXT files. 




  

    

  

  





  Key Features •

  How To Use •

  Requirements •

  Copyright









## Key Features

* File handling

  - The script supports PDF, DOCX, and TXT files

  - Read the content using the pdfplumber, docx, and built-in open() functions

* GPT-3 integration

  - The script uses the OpenAI GPT-3 model, specifically the `text-davinci-003` engine, to generate answers to questions.

* Confidence scoring

  - The script calculates confidence scores for the generated answers using log probabilities returned by the GPT-3 API.

* Concurrency

  - It uses the `concurrent.futures.ThreadPoolExecutor` to process questions concurrently, potentially speeding up the process.

* Text preprocessing

  - The script splits the input document into chunks to fit within GPT-3's token limit, and post-processes the answers to remove duplicate sentences.

* Saving conversation history

  - The script allows users to save the conversation history to a text file.

* Caching

  - The script uses `lru_cache` decorator to cache the answers generated by GPT-3. This way, if a user asks the same question again, the cached answer can be returned instead of making another API call.

* Gui

  - The script provides a friendly graphical user interface built using the tkinter library and `ttkthemes` allowing users to select a file, input a question, view the answer, and save the conversation history.

## How To Use

- Put you api key in line 45

- Run the script

- Select your file

- Enter your question and click submit

It's as simple as that

> **Note**

> We will provide an executable version soon

## Requirements

`pip install openai pdfplumber python-docx`

## Copyright

All rights reserved to Bropocalypse Team.