In this project, I develop a basic chatbot using the llama-index package and Streamlit. Check out the demo here.

My aim is to get hands-on experience with the llama-index package and to test the chatbot's functionality in a simple use case scenario: answering questions about a CV.

Use Case

I've included a single CV document for simplicity. The chatbot is designed to respond to questions based on this CV. This setup allows me to evaluate both the chat functionality and the llama-index package efficiently. Feel free to try out the demo and see how the chatbot handles questions about the provided CV.

Here's a brief overview drawing of the project:

System Overview

The chatbot uses the llama index package to answer questions. The package is used to retrieve relevant documents from the llama index and then generate a response using the retrieved documents. Here is a high-level overview of the system:

Steps

The system goes through the following steps:

• User add documents

  • The user adds their documents (.txt files) to data folder.
  • The system show the user the documents added.

• Indexing

  • The system processes the uploaded documents and creates an index.

• Query Engine Development

  • The system develops a query engine, which is a component responsible for interpreting user queries and searching through the document index to retrieve relevant documents.

• Query Processing

  • a When the user asks a question, the chatbot utilizes the query engine to process the query.

UI

The following screenshot shows the user interface of the application.

Setup and Usage

It's recommended to create a virtual environment. Here, we'll be using Conda. To create a new Conda environment, use the following command:

conda create --name llm

After creating the environment, activate it using:

conda activate llm

Once the Conda environment is activated, you can install the dependencies from the requirements.txt file. Use the following command:

pip install -r requirements.txt

To run the code, use the following command:

streamlit run main.py