QA Bot for DataTalks.Club Zoomcamp Courses

This project implements a Question-Answering (QA) Bot designed to assist participants in completing DataTalks.Club Zoomcamp courses (data engineering, ML engineering, & MLOps) The bot leverages a RAG design, using course FAQ documents indexed with ElasticSearch for retrieval & (for now) OpenAI's GPT-4o for response generation.

Features

• Document Retrieval: Fetches relevant documents based on the user's query.
• Filtering by Course Content: Allows filtering the documents by specific course sections for more accurate answers.
• Customized Prompts: Builds prompts for OpenAI's API using the retrieved documents, ensuring the answers are contextually relevant.
• OpenAI Integration: Utilizes OpenAI's API to generate answers to user queries.

Getting Started

Prerequisites

• Python 3.8 or higher
• Pipenv for dependency management
• OpenAI API key

Installation

1. Clone the repository:

   git clone https://github.com/a-kuhn/llm-zoomcamp.git

2. Navigate to the project directory: cd llm-zoomcamp

3. Install dependencies using Pipenv: pipenv install

4. Activate the virtual environment: pipenv shell

Configuration

Create a .env file in the project root directory and add your OpenAI API key: OPENAI_API_KEY='your_openai_api_key'

Running the Application

Start the ElasticSearch container

must first create a volume & index the FAQ docs (see jupyter notebook)
> ./elasticsearch/scripts/run_elasticsearch_w_volumes.sh

Execute the main script to start the QA Bot:

> python3 -m api.app -q "how to run docker" -f "machine-learning-zoomcamp"