Streamlit Chatbot with Elasticsearch and Hugging Face

This is a simple chatbot application built using Streamlit. The application uses Elasticsearch for document retrieval and Hugging Face's transformers for generating responses.

Features

• Multimodal RAG: Combines document retrieval and response generation.
• Elasticsearch: Retrieves relevant documents based on user queries.
• Hugging Face: Generates responses using the GPT-2 model or other Hugging Face models.
• Streamlit Interface: Provides a simple and interactive web interface.

Prerequisites

• Python 3.7 or higher
• Elasticsearch server (local or hosted)
• Hugging Face API key

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/streamlit-chatbot.git
   cd streamlit-chatbot

2. Install the required packages:

   pip install -r requirements.txt

3. Ensure your Elasticsearch server is running and has an index named documents.

Usage

1. How to run it on your own machine

Install the requirements

$ pip install -r requirements.txt

Run the app

$ streamlit run streamlit_app.py

2. Open your browser and go to http://localhost:8501.

3. Enter your Elasticsearch server URL and Hugging Face API key when prompted.

Elasticsearch Setup

1. Install Elasticsearch (if not already installed):

   • For local installation, download from Elastic's website and follow the instructions.
   • For a hosted service, consider using Elasticsearch Service.

2. Start your Elasticsearch server:

   ./bin/elasticsearch

3. Create an index and add documents:

   # Create an index named 'documents'
   curl -X PUT "localhost:9200/documents"
   
   # Add a document to the index
   curl -X POST "localhost:9200/documents/_doc/1" -H 'Content-Type: application/json' -d'
   {
     "content": "This is the content of the document."
   }
   '

Example Document

Ensure your documents in Elasticsearch have the following structure:

{
  "content": "This is the content of the document."
}