Build Your Own AI Chatbot

Background

• Retrieval Augmented Generation (RAG) for chatbots.

Chatbot components - Overview and create accounts

• Large Language Model: Google Gemini.
• User Interface: Streamlit.
• Connection: LlamaIndex
• Version control/code hosting: Github

Build our chatbot!

Customization Discussion

• What chatbot would you like to create?

Walking through changes

• Changing documents
• Changing the prompt
• Other changes to consider

Questions and further discussion

RAG - Retrieval Augmented Generation

• Natural language processing (NLP) is a technique that combines retrieval and generative-based approaches to improve the quality and relevance of generated text.
• To respond to user questions, it uses a "vector database" to retrieve relevant information from a database or documents + a large language model.
• It reduces the chance for the AI to hallucinate (generate false or synthetic information)

KingbotGPT, SJSU's Library Chatbot

• Python-bot created using Streamlit, LangChain, and GPT 40 mini.
• Team:
  • Jessie Cai, Nick Szydlowski, Sharesly Rodriguez, Sudip Das, Suhaas Teja Vijjagiri, Saneeth Reddy Chimmula (2023-2024).
• Retrieval Augmented Generation.
  • Vector Database: Local Website Crawl.
    • Library Research Guides.
    • Library Website.

Create your own example lets you:

• Try out different data in the RAG workflow.
• Try different default prompts.

Components of Our Chatbot