📜Docs-ai-chatbot

Table of Contents

• First look
• Installation
• Model
• App appearance

First look 👀

An application written in Python allows us to chat with our PDF and CSV files.

By using the LLM GPT4ALL module from Langchain, we read data from the file. Then the chatbot uses the Mistral 7B model to return answers to the questions asked.

The user interface is based on the open-source Streamlit library.

Installation ℹ

• First of all, clone this repository on your device. Open your shell and paste the following commands:

URL cloning

git clone https://github.com/maciekmalachowski/Docs-ai-chatbot.git

SSH cloning

git clone git@github.com:maciekmalachowski/Docs-ai-chatbot.git

• Then you need to install the requirements contained in requirements.txt.

pip install -r requirements.txt

• After that run internal streamlit server in the repository folder.

streamlit run main.py

Streamlit will launch the main.py file contained in the folder and automatically display it in the browser.

Model 📂

Unfortunately, the model is too large to upload to this project, so you have to download it yourself.

It is very simple. Just go to the GPT4ALL and choose the model that suits you best.

In this project, I am using mistral-7b-instruct-v0.1.Q4_0.gguf.

Then place it in the docs/model folder.

App appearance 🔍

The front page doesn't have much, but that was the idea - to make the application trivially easy to use.

After uploading the file and pressing the Process button, a window will appear where we can ask questions about the document.

Additionally, when you upload a CSV file, a table with data from that document will be displayed.

As the name of the project suggests, questions and answers are displayed in the form of a chat.