MLFF_QD Expert Chatbot 🤖

A RAG (Retrieval-Augmented Generation) chatbot designed to answer questions about MLFF_QD, configurations, and workflows.

🌼 Features

• Hybrid Search: Combines ChromaDB (Vector) and BM25 (Keyword) for robust retrieval.
• Re-Ranking: Uses FlashRank to re-order retrieved documents for higher relevance.
• LLM Integration: Powered by Groq (Llama 3) for fast and accurate responses.
• Interactive UI: Built with Chainlit for a chat-like experience.
• Sources Citation: Displays the specific files used to generate the answer.

📁 Project Structure

data/                  # Place your source documents here
│   manual_user_guide.txt
│   paper.docx
│   qa_pairs_new.txt
app_chainlitkeywordRank.py   # Main application logic
config.py             # Configuration settings
.env                  # Environment variables (API Keys)
requirements.txt      # Python dependencies

🛠 Installation

Prerequisites

• Anaconda or Miniconda
• Git

1. Clone the Repository

git clone https://github.com/nlesc-nano/MLFF_QD_ChatBot
cd MLFF_QD_ChatBot

2. Create a Conda Environment

conda create -n mlff_bot python=3.11 -y
conda activate mlff_bot

3. Install Dependencies

pip install -r requirements.txt

4. Configure Environment Variables

1. Create a file named .env in the root directory.
2. Add your Groq API Key:

GROQ_API_KEY=gsk_your_actual_api_key_here

5. Prepare Data

Ensure your data files are in the data/ folder. The app expects:

• manual_user_guide.txt
• qa_pairs_new.txt
• paper.docx

(You can change these filenames in config.py if needed.)

⚠️ Important Note on File Formats

The current code supports .txt and .docx files only.
If you want to use other file formats (e.g., PDF, CSV), you must modify the setup_retriever
function inside app_chainlitkeywordRank.py to include the appropriate document loader.

You can find the list of supported loaders here:
👉 LangChain Document Loaders Documentation

🚀 Usage

Run the Chainlit application:

chainlit run app_chainlitkeywordRank.py -w

The -w flag enables auto-reloading when you change code.