Chat-Based Question Answering System with LangChain and Transformers

This project is a simple implementation of a Question Answering System built using LangChain and Transformers. It utilizes document embedding, vector search, and a question-answering pipeline to extract answers from a given text document.

Features

• Document Parsing: Loads a text document and splits it into manageable chunks.
• Vector Search: Uses FAISS for efficient similarity search within the document.
• Embeddings: Employs HuggingFace's sentence-transformers/all-MiniLM-L6-v2 for text embeddings.
• Question Answering: Utilizes distilbert-base-cased-distilled-squad for answering questions based on the retrieved context.
• Extensible Configuration: Customize questions and documents easily through the Config class.

---

Installation

1. Clone this repository:

   git clone https://github.com/Niewiaro/AI-RAG.git
   cd AI-RAG

2. Install required dependencies:

   pip install -r requirements.txt

---

Usage

Configuration

Edit the Config class in the script to set your custom document and questions:

class Config:
    def __init__(self):
        self.questions = (
            "Who am I writing with?",
            "When will he arrive?",
            "Tell me his favorite song",
        )
        self.file_name = "document.txt"
        self.document = """
        Your document content here...
        """

Running the Script

1. Run the script:

   python main.py

2. The script will:

   • Save the document as a text file.
   • Build a vector store from the document.
   • Execute the specified questions.
   • Print the answers to the console.

---

Example Output

For the given example document and questions, the output might look like this:

Q: Who am I writing with?
A: Michał

Q: When will he arrive?
A: 5:55 PM

Q: Tell me his favorite song
A: Mateusz, flip the switch

---

Project Structure

• main.py: The main script containing the logic for loading documents, building a vector store, and executing the question-answering pipeline.
• document.txt: The sample document file created by the script.

---

Technologies Used

• LangChain: Framework for working with language models and text embeddings.
• Transformers: HuggingFace's library for NLP models.
• FAISS: Library for efficient similarity search.

---

Future Enhancements

• Add support for more languages using multilingual models.
• Improve accuracy by fine-tuning the QA model on specific domains.
• Integrate with a front-end interface for user-friendly interaction.

---

Acknowledgements

Special thanks to:

• HuggingFace for providing state-of-the-art models.
• LangChain for making it easier to work with LLMs and embeddings.