CoolFace
Apppublic

antonsoloviev/ecommerce-support-chatbot

sourceHugging Facemitupdated 2y agoView on Hugging Face
0likes
App README

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

E-commerce Customer Support Chatbot with Retrieval-Augmented Generation (RAG)

This project implements an intelligent customer support chatbot for an e-commerce store using Retrieval-Augmented Generation (RAG). The chatbot leverages a pre-trained Large Language Model (LLM) combined with a vector database of product information and frequently asked questions (FAQs) to provide accurate and helpful answers to user queries.

Project Description

The chatbot is designed to enhance the customer support experience by providing instant answers to common questions about products, shipping, returns, and other store policies. It utilizes the following core components:

  • —Retrieval-Augmented Generation (RAG): Combines the power of a pre-trained LLM with a specialized knowledge base to retrieve relevant information and generate contextually appropriate responses.
  • —Vector Database (FAISS): Stores embeddings of product information and FAQs, enabling efficient similarity search to find the most relevant context for a given query.
  • —Large Language Model (LLM): Uses the Mistral-7B-Instruct-v0.2 model from Hugging Face Transformers to generate human-like responses based on the retrieved context and the user's question.
  • —Sentence Transformers: Employs the sentence-transformers/all-mpnet-base-v2 model to create high-quality sentence embeddings for efficient information retrieval.
  • —Streamlit: A user-friendly web application framework for building and deploying the chatbot interface.
  • —Hugging Face Spaces: The chatbot is deployed on Hugging Face Spaces for easy accessibility and demonstration.

Demo

Experience the live chatbot demo here: https://huggingface.co/spaces/antonsoloviev/ecommerce-support-chatbot

Dataset

This project utilizes the AmazonQA dataset, a large-scale collection of question-answer pairs related to Amazon products.

Dataset Source: https://github.com/amazon-research/open-question-answering-data/tree/main/AmazonQA

Download Instructions:

The dataset is not included in this repository due to its size. Please follow the instructions in data/HOW_TO_DOWNLOAD_DATA.md to download the dataset.

Project Structure

ecommerce-support-chatbot/
├── data/
│ └── AmazonQA/ (Place the downloaded AmazonQA dataset files here)
│ └── faq/ (Optional: Add any additional FAQ documents here)
│ └── HOW_TO_DOWNLOAD_DATA.md (Instructions for downloading the dataset)
├── model/ (Directory for storing the downloaded LLM and sentence-transformer model)
├── vectorstore/ (Directory for storing the FAISS index)
├── notebooks/
│ └── chatbot_development.ipynb (Colab notebook for development and experimentation)
├── src/
│ ├── app.py (Streamlit application script)
│ ├── train.py (Script for creating the vector store)
│ └── download_data.py (Script to download the AmazonQA dataset)
├── README.md
├── requirements.txt
└── .gitignore

Installation

  1. 1.Clone the repository:
bash
    git clone https://github.com/antonsoo/ecommerce-support-chatbot
    cd ecommerce-support-chatbot
  1. 1.Create a virtual environment (recommended):
bash
    python3 -m venv .venv
    source .venv/bin/activate  # On Linux/macOS
    .venv\Scripts\activate  # On Windows
  1. 1.Install dependencies:
bash
    pip install -r requirements.txt
  1. 1.Download the dataset:

Follow the instructions in data/HOW_TO_DOWNLOAD_DATA.md to download the AmazonQA dataset and place it in the data/AmazonQA directory.

Usage

To create the vector store:

  1. 1.Make sure you have downloaded the dataset and placed it in the correct directory.
  2. 2.Run the training script:
bash
    python src/train.py

This will create the FAISS vector store from the dataset and save it to the vectorstore directory.

To run the Streamlit app locally:

  1. 1.Ensure that the vector store has been created by running train.py.
  2. 2.Run the following command:
bash
    streamlit run src/app.py

This will start the Streamlit app, and you will be able to access it in your web browser (usually at http://localhost:8501).

To test locally using ngrok (Optional):

  1. 1.Install ngrok following instructions here
  2. 2.Update the NGROK_AUTH_TOKEN in the run_ngrok() function in the test_locally.py file.
  3. 3.Run the test_locally.py script to start Streamlit and ngrok in separate threads:
bash
    python src/test_locally.py

This will start Streamlit in the background, start ngrok, and provide a temporary public URL to access the app. This is useful for testing in restricted environments like Google Colab.

Deployment:

The app is deployed on Hugging Face Spaces. To deploy your own version:

  1. 1.Create a Hugging Face account and a new Space, selecting "Streamlit" as the SDK.
  2. 2.Connect your GitHub repository to the Space.
  3. 3.Add your Hugging Face token as a secret named HUGGINGFACE_TOKEN in the Space's settings.
  4. 4.Hugging Face Spaces will automatically build and deploy your app.

Model Details

  • —LLM: mistralai/Mistral-7B-Instruct-v0.2
  • —Embedding Model: sentence-transformers/all-mpnet-base-v2
  • —Vector Store: FAISS

Ethical Considerations

  • —This project uses a publicly available dataset for research and educational purposes.
  • —It is important to be aware of potential biases in the data and to use the chatbot responsibly.
  • —The chatbot is intended to be a helpful tool, but it should not be considered a replacement for human customer support agents.

Future Improvements

  • —Experiment with different LLMs and embedding models.
  • —Fine-tune the LLM on a domain-specific dataset.
  • —Incorporate a mechanism for handling out-of-domain or unsupported questions.
  • —Improve the user interface and add more interactive features to the Streamlit app.
  • —Evaluate the chatbot's performance using more comprehensive metrics and user studies.

Acknowledgements

  • —Hugging Face for the Transformers library, pre-trained models, and the Spaces platform.
  • —Streamlit for the web app framework.
  • —LangChain for the RAG implementation.
  • —The creators of the AmazonQA dataset.

License

This project is licensed under the MIT License - see the LICENSE file for details.