Sun Sun Chatbot - Question Answering System

Project Description

This project is for AI R&D Challenge for Sun* Summer Internship

Author: Nguyen Tuan Ngoc, Sophomore from VNU-UET at this time

This project is my Question-Answering System, you can ask anything. To improve the quality of the answer, you can add your data file into vector database to improve the quality with private data.

🚀 Demo

link

🖼️ Project Screenshots:

🧐 Features

Here're some of the project's best features:

• Chatting with Chatbot to help you address your problems.
• You can add data with doc, docx(error), pdf format to enhance your database and improve the quality of the answer which based on RAG architecture.

🛠️ Installation Steps:

1. Privileges

Python Version 3.10 
Conda Command Prompt

2. Conda Setup

conda create -n agent python=3.10 -y 
conda activate agent
#using pip to install all package from requirements.txt file 
pip install -r requirements.txt

💻 Built with

Technologies used in the project:

• Frontend:
  • Streamlit
• Backend:
  • Langchain
  • Zenguard for Prompt Injection Preserving (Not done)
  • FastAPI
• Vector database:
  • ChromaDB
• RAG Techniques:
  • Advanced RAG ReAct Agent
  • Hybrid Search
  • Agent External Tool: BingSearch Tool.
  • Semantic Chunking
  • Contextual Compression
  • ReRanker (Demo Colab)
• LLM API:
  • OpenAI API Key

🔗 Pipeline

📁 Project Structure

* Tools:
    * AgentTools
        - BingSearch.py | Bing Search Tools
        - RAG.py | RAG Tools
        - Tools.py | Various Tools Powered By Agent
    * Retrieval:
        - Model.py | Model For Retrieval
        - Retrieval.py | Retrieval Tools
    * VectorDatabase:
        - API.py | API For Vector Database
        - Chroma.py | Vector Database Tools
* Config_Model.yaml (use for higher performance configuration)
* Config.py
* RunBackend.py
* requirements.txt
* frontend:
    * app.py: Streamlit Server
    * client.py
* RAG.py: Classical RAG with Vector Database

🗎 API docs

• In Tools\VectorDatabase\API.py:
  • add_data: upload pdf file or doc file (still error) to chunk and store in Chroma database.
  • delete_database: delete database
• In RunBackend.py:
  • query_handler: use for process questions and generate answers.
• In RAG.py:
  • query_handler: use for process questions and generate answers.

👨🏻‍💻 How to run this project

1. Clone this project from GitHub

git clone https://github.com/ngoctuannguyen/SunSunChatBot.git

2. Run Backend Server

python RunBackend.py (Question Answering Server)
python Tools\VectorDatabase\API.py (File Upload Server)

3. Run Frontend Server

streamlit run frontend/app.py

4. Run Classical RAG (due to lackage of data)

Modify the RAG.py: change result of def get_response() to response.json()["text"]

👨🏻 How to use this project

Try some questions in test.txt or your own questions

📩 Contact

If you have questions, please issue this repository or contact me via email [email protected].