ATO Chatbot

A RAG-based chatbot system for Australian Taxation Office (ATO) information retrieval and assistance, powered by data sourced from ato.gov.au.

Live Demo

https://ato-chat.streamlit.app/

Figure 1: Streamlit Chat Interface with example conversation

Overview

This project implements a Retrieval-Augmented Generation (RAG) chatbot system specifically designed for ATO-related queries. It consists of two main components:

1. Data Pipeline & Model Training: A modular pipeline built with ZenML for data processing and index creation
2. Interactive Interface: A Streamlit-based chat interface for user interactions

Architecture

Technology Stack

• Data Collection & Processing

  • Firecrawl - Web crawling and content extraction
  • ZenML - MLOps pipeline orchestration
  • Qdrant - Vector database for embeddings storage

• Machine Learning & AI

  • OpenAI - Large Language Model API
  • LlamaIndex - RAG framework and indexing

• Backend & Infrastructure

  • Docker - Containerization
  • MongoDB - Document storage

• Frontend

  • Streamlit - Interactive web interface
  • Streamlit-Chat - Chat UI components

Components

1. Data Pipeline

The data pipeline is built using ZenML and consists of several key steps:

1. Data Collection: Uses Firecrawl to extract content from ATO pages
2. Data Cleaning: Processes and filters the collected data
3. Index Creation: Creates embeddings and stores them in Qdrant

Figure 2: ZenML Pipeline Workflow showing data processing steps

Key pipeline components:

python:src/ato_chatbot/pipelines/simple_index_pipeline.py

2. Chat Interface

The chat interface is built with Streamlit and implements a 3-step RAG process:

1. Query Rephrasing: Improves query understanding
2. Knowledge Retrieval: Fetches relevant information from Qdrant
3. Response Generation: Uses OpenAI to generate contextual responses

Key interface components:

python:src/ato_chatbot/chat_interface.py

Setup

Prerequisites

• Python 3.12+g
• Docker and Docker Compose
• OpenAI API key

Installation

1. Clone the repository
2. Install dependencies:

uv install

3. Start required services:

make up

Running the Application

1. Train the model:

make zen_run_simple_index

2. Start the chat interface:

make streamlit

Dependencies

Key dependencies include:

toml:pyproject.toml

License

Apache License 2.0