A versatile AI-powered chatbot built with Streamlit, integrating multiple LLMs for chat, web search, and file-based Q&A.
MIT License
This app provides three main features: a standard Chatbot for general conversations, a more advanced SearchBot that leverages Large Language Models (LLMs) with search tools, and a File Q&A feature that allows you to interact with documents by asking questions based on their content.
Explore MultiAI at: https://multi-ai.streamlit.app/
This is just a regular chatbot where you can interact with various open source Large Language Models (LLMs). Whether you want to ask questions, chat about random topics, or just explore what these models can do, this chatbot is here to help. It's a fun and simple way to see how LLMs handle different kinds of conversations.
SearchBot is a bit more advanced. It combines the power of LLMs with search tools, like DuckDuckGo's LangChain Python modules, to give you more accurate answers. The LLMs generate search queries, and then the results are used to build responses. Not all LLMs are great at using these search tools, so you might see some errors occasionally. Below is a list of LLMs that most consistently perform well with the search tool:
With the File Q&A feature, you can upload files in PDF, TXT, or Markdown (MD) formats and ask questions based on the content of those files. The LLMs use Retrieval-Augmented Generation (RAG), which means they retrieve relevant information from the document and generate answers that are contextually accurate. This approach helps the model provide more precise responses based on the actual content of the files, making it a handy way to interact with your documents without having to read through everything yourself.
ZERO_SHOT_REACT_DESCRIPTION
agents with web search tools.all-MiniLM-L12-v2
for embedding text data in the File Q&A feature.requirements.txt
)git clone https://github.com/BilalM04/MultiAI.git
cd MultiAI
pip install -r requirements.txt
streamlit run Chatbot.py
localhost
link specified in the command line.Explore the live demo: MultiAI